{"owner":"coreyhaines31","repo":"marketingskills","hasSkills":true,"totalSkillsCount":389,"totalTokensCount":875545,"categories":["marketplace","plugin-manifest","root-instruction","claude-rule","anthropic-skill"],"hasMcp":false,"mcpConfig":null,"found":[".agents/plugins/marketplace.json",".claude-plugin/marketplace.json",".claude-plugin/plugin.json",".codex-plugin/plugin.json","AGENTS.md","CLAUDE.md","skills/ab-testing/SKILL.md","skills/ab-testing/evals/evals.json","skills/ab-testing/references/sample-size-guide.md","skills/ab-testing/references/test-templates.md","skills/ad-creative/SKILL.md","skills/ad-creative/evals/evals.json","skills/ad-creative/references/awareness-headlines.md","skills/ad-creative/references/creative-review-page.md","skills/ad-creative/references/creative-roadmap.md","skills/ad-creative/references/generative-tools.md","skills/ad-creative/references/hook-system.md","skills/ad-creative/references/imessage-video-ads.md","skills/ad-creative/references/meta-creative-formats.md","skills/ad-creative/references/motion-video-ads.md","skills/ad-creative/references/platform-specs.md","skills/ad-creative/references/short-form-video-specs.md","skills/ad-creative/references/static-ad-templates.md","skills/ads/SKILL.md","skills/ads/evals/evals.json","skills/ads/references/abm-playbook.md","skills/ads/references/ad-copy-templates.md","skills/ads/references/audience-targeting.md","skills/ads/references/audit-guardrails.md","skills/ads/references/b2b-paid-playbook.md","skills/ads/references/creative-research-automation.md","skills/ads/references/google-ads-audit-checklist.md","skills/ads/references/google-search-playbook.md","skills/ads/references/linkedin-b2b-playbook.md","skills/ads/references/meta-andromeda-playbook.md","skills/ads/references/meta-decision-system.md","skills/ads/references/operating-check.md","skills/ads/references/payback-period.md","skills/ads/references/platform-setup-checklists.md","skills/ads/references/reading-google-ads-data.md","skills/ads/references/rsa-output-spec.md","skills/ai-seo/SKILL.md","skills/ai-seo/evals/evals.json","skills/ai-seo/references/agent-defaults.md","skills/ai-seo/references/agent-readiness.md","skills/ai-seo/references/ai-attribution.md","skills/ai-seo/references/answer-accuracy.md","skills/ai-seo/references/citations-vs-recommendations.md","skills/ai-seo/references/content-patterns.md","skills/ai-seo/references/content-types.md","skills/ai-seo/references/format-volatility.md","skills/ai-seo/references/linkedin-ai-citations.md","skills/ai-seo/references/multilingual-and-commerce.md","skills/ai-seo/references/okf.md","skills/ai-seo/references/platform-ranking-factors.md","skills/ai-seo/references/positioning-and-consensus.md","skills/ai-seo/references/prompt-portfolio.md","skills/ai-seo/references/youtube-ai-citations.md","skills/analytics/SKILL.md","skills/analytics/evals/evals.json","skills/analytics/references/event-library.md","skills/analytics/references/ga4-implementation.md","skills/analytics/references/gtm-implementation.md","skills/aso/SKILL.md","skills/aso/evals/evals.json","skills/aso/references/apple-specs.md","skills/aso/references/benchmarks.md","skills/aso/references/google-play-specs.md","skills/aso/references/report-template.md","skills/aso/references/scoring-criteria.md","skills/attribution/SKILL.md","skills/attribution/evals/evals.json","skills/attribution/references/attribution-models.md","skills/attribution/references/by-business-type.md","skills/attribution/references/first-party-tracking.md","skills/attribution/references/measurement-paradigms.md","skills/churn-prevention/SKILL.md","skills/churn-prevention/evals/evals.json","skills/churn-prevention/references/cancel-flow-patterns.md","skills/churn-prevention/references/dunning-playbook.md","skills/co-marketing/SKILL.md","skills/co-marketing/evals/evals.json","skills/co-marketing/references/partnership-types.md","skills/cold-email/SKILL.md","skills/cold-email/evals/evals.json","skills/cold-email/references/benchmarks.md","skills/cold-email/references/deliverability.md","skills/cold-email/references/follow-up-sequences.md","skills/cold-email/references/frameworks.md","skills/cold-email/references/linkedin-outreach.md","skills/cold-email/references/multichannel-cadence.md","skills/cold-email/references/outbound-plays.md","skills/cold-email/references/personalization.md","skills/cold-email/references/reply-handling.md","skills/cold-email/references/subject-lines.md","skills/community-marketing/SKILL.md","skills/community-marketing/evals/evals.json","skills/community-marketing/references/community-models.md","skills/competitor-profiling/SKILL.md","skills/competitor-profiling/evals/evals.json","skills/competitor-profiling/references/templates.md","skills/competitor-profiling/references/tool-reference.md","skills/competitors/SKILL.md","skills/competitors/evals/evals.json","skills/competitors/references/content-architecture.md","skills/competitors/references/templates.md","skills/content-strategy/SKILL.md","skills/content-strategy/evals/evals.json","skills/content-strategy/references/content-distribution.md","skills/content-strategy/references/headless-cms.md","skills/conversion-tracking/SKILL.md","skills/conversion-tracking/assets/gtm-recipes/README.md","skills/conversion-tracking/assets/gtm-recipes/detect/detect-acuity-scheduling.json","skills/conversion-tracking/assets/gtm-recipes/detect/detect-cal-com.json","skills/conversion-tracking/assets/gtm-recipes/detect/detect-calendly.json","skills/conversion-tracking/assets/gtm-recipes/detect/detect-contact-form-7.json","skills/conversion-tracking/assets/gtm-recipes/detect/detect-divi-forms.json","skills/conversion-tracking/assets/gtm-recipes/detect/detect-elementor-forms.json","skills/conversion-tracking/assets/gtm-recipes/detect/detect-file-download.json","skills/conversion-tracking/assets/gtm-recipes/detect/detect-fluent-forms.json","skills/conversion-tracking/assets/gtm-recipes/detect/detect-formidable-forms.json","skills/conversion-tracking/assets/gtm-recipes/detect/detect-forminator.json","skills/conversion-tracking/assets/gtm-recipes/detect/detect-framer-forms.json","skills/conversion-tracking/assets/gtm-recipes/detect/detect-generic-ajax-form.json","skills/conversion-tracking/assets/gtm-recipes/detect/detect-gravity-forms.json","skills/conversion-tracking/assets/gtm-recipes/detect/detect-hubspot-forms.json","skills/conversion-tracking/assets/gtm-recipes/detect/detect-jotform.json","skills/conversion-tracking/assets/gtm-recipes/detect/detect-ninja-forms.json","skills/conversion-tracking/assets/gtm-recipes/detect/detect-phone-click.json","skills/conversion-tracking/assets/gtm-recipes/detect/detect-squarespace-forms.json","skills/conversion-tracking/assets/gtm-recipes/detect/detect-tally.json","skills/conversion-tracking/assets/gtm-recipes/detect/detect-thank-you-page.json","skills/conversion-tracking/assets/gtm-recipes/detect/detect-typeform.json","skills/conversion-tracking/assets/gtm-recipes/detect/detect-webflow-forms.json","skills/conversion-tracking/assets/gtm-recipes/detect/detect-wix-forms.json","skills/conversion-tracking/assets/gtm-recipes/detect/detect-wix-scheduling.json","skills/conversion-tracking/assets/gtm-recipes/detect/detect-wpforms.json","skills/conversion-tracking/assets/gtm-recipes/detect/detect-ws-form.json","skills/conversion-tracking/assets/gtm-recipes/event-map.json","skills/conversion-tracking/assets/gtm-recipes/send/ga4.json","skills/conversion-tracking/assets/gtm-recipes/send/google-ads.json","skills/conversion-tracking/assets/gtm-recipes/send/linkedin.json","skills/conversion-tracking/assets/gtm-recipes/send/meta.json","skills/conversion-tracking/assets/gtm-recipes/send/microsoft.json","skills/conversion-tracking/assets/gtm-recipes/send/tiktok.json","skills/conversion-tracking/assets/snippets/README.md","skills/conversion-tracking/evals/evals.json","skills/conversion-tracking/references/audit-playbook.md","skills/conversion-tracking/references/browser-side.md","skills/conversion-tracking/references/conversion-moments.md","skills/conversion-tracking/references/discrepancies.md","skills/conversion-tracking/references/ecommerce-platforms.md","skills/conversion-tracking/references/google-ads.md","skills/conversion-tracking/references/meta.md","skills/conversion-tracking/references/other-platforms.md","skills/conversion-tracking/references/server-side.md","skills/copy-editing/SKILL.md","skills/copy-editing/evals/evals.json","skills/copy-editing/references/checklist.md","skills/copy-editing/references/content-refresh.md","skills/copy-editing/references/plain-english-alternatives.md","skills/copywriting/SKILL.md","skills/copywriting/evals/evals.json","skills/copywriting/references/ai-tells.md","skills/copywriting/references/copy-frameworks.md","skills/copywriting/references/natural-transitions.md","skills/cro/SKILL.md","skills/cro/evals/evals.json","skills/cro/references/experiments.md","skills/cro/references/form.md","skills/customer-research/SKILL.md","skills/customer-research/evals/evals.json","skills/customer-research/references/interviews-and-surveys.md","skills/customer-research/references/source-guides.md","skills/directory-submissions/SKILL.md","skills/directory-submissions/evals/evals.json","skills/directory-submissions/references/directory-list.md","skills/directory-submissions/references/positioning-variations.md","skills/emails/SKILL.md","skills/emails/evals/evals.json","skills/emails/references/copy-guidelines.md","skills/emails/references/email-types.md","skills/emails/references/sequence-templates.md","skills/events/SKILL.md","skills/events/evals/evals.json","skills/events/references/event-portfolio-strategy.md","skills/events/references/speaking.md","skills/events/references/sponsorship-roi.md","skills/events/references/webinar-funnel.md","skills/free-tools/SKILL.md","skills/free-tools/evals/evals.json","skills/free-tools/references/tool-benchmarks.md","skills/free-tools/references/tool-types.md","skills/image/SKILL.md","skills/image/evals/evals.json","skills/image/references/ai-image-prompting.md","skills/influencer-marketing/SKILL.md","skills/influencer-marketing/evals/evals.json","skills/influencer-marketing/references/newsletter-sponsorships.md","skills/influencer-marketing/references/paid-ads-creator-program.md","skills/influencer-marketing/references/telegram-channels.md","skills/influencer-marketing/references/ugc-creator-program.md","skills/launch/SKILL.md","skills/launch/evals/evals.json","skills/launch/references/shipped-changes.md","skills/launch/references/site-launch-qa.md","skills/lead-magnets/SKILL.md","skills/lead-magnets/evals/evals.json","skills/lead-magnets/references/benchmarks.md","skills/lead-magnets/references/content-to-skill-corpus-design.md","skills/lead-magnets/references/content-to-skill-release.md","skills/lead-magnets/references/content-to-skill.md","skills/lead-magnets/references/format-guide.md","skills/link-building/SKILL.md","skills/link-building/evals/evals.json","skills/link-building/references/prospecting-and-outreach.md","skills/link-building/references/risk-policy.md","skills/link-building/references/tactics.md","skills/marketing-council/SKILL.md","skills/marketing-council/evals/evals.json","skills/marketing-council/references/advisor-template.md","skills/marketing-council/references/advisors/alex-hormozi.md","skills/marketing-council/references/advisors/ann-handley.md","skills/marketing-council/references/advisors/april-dunford.md","skills/marketing-council/references/advisors/byron-sharp.md","skills/marketing-council/references/advisors/claude-hopkins.md","skills/marketing-council/references/advisors/david-ogilvy.md","skills/marketing-council/references/advisors/eugene-schwartz.md","skills/marketing-council/references/advisors/gary-halbert.md","skills/marketing-council/references/advisors/gary-vaynerchuk.md","skills/marketing-council/references/advisors/rory-sutherland.md","skills/marketing-council/references/advisors/russell-brunson.md","skills/marketing-council/references/advisors/seth-godin.md","skills/marketing-ideas/SKILL.md","skills/marketing-ideas/evals/evals.json","skills/marketing-ideas/references/guerrilla-marketing.md","skills/marketing-ideas/references/ideas-by-category.md","skills/marketing-loops/SKILL.md","skills/marketing-loops/evals/evals.json","skills/marketing-loops/references/loop-catalog.md","skills/marketing-loops/references/loop-guardrails.md","skills/marketing-loops/references/loop-orchestration.md","skills/marketing-loops/references/loop-state.md","skills/marketing-loops/references/loop-template.md","skills/marketing-loops/references/outbound-operator.md","skills/marketing-loops/references/seo-operator.md","skills/marketing-plan/SKILL.md","skills/marketing-plan/evals/evals.json","skills/marketing-plan/references/aarrr-framework.md","skills/marketing-plan/references/budget-planning.md","skills/marketing-plan/references/client-types.md","skills/marketing-plan/references/current-state-rubric.md","skills/marketing-plan/references/example-quietude.md","skills/marketing-plan/references/funding-stage-unlocks.md","skills/marketing-plan/references/growth-patterns.md","skills/marketing-plan/references/idea-cross-reference.md","skills/marketing-plan/references/market-sizing.md","skills/marketing-plan/references/measurement-framework.md","skills/marketing-plan/references/methodology.md","skills/marketing-plan/references/ops-stack-mapping.md","skills/marketing-plan/references/plan-template.md","skills/marketing-plan/references/team-and-agency-model.md","skills/marketing-psychology/SKILL.md","skills/marketing-psychology/evals/evals.json","skills/offers/SKILL.md","skills/offers/evals/evals.json","skills/offers/references/bonus-stacking.md","skills/offers/references/examples.md","skills/offers/references/guarantee-design.md","skills/offers/references/offer-anatomy.md","skills/offers/references/offer-formats.md","skills/offers/references/saas-offers.md","skills/offers/references/scarcity-urgency.md","skills/offers/references/value-equation.md","skills/onboarding/SKILL.md","skills/onboarding/evals/evals.json","skills/onboarding/references/activation-models.md","skills/onboarding/references/experiments.md","skills/onboarding/references/minimum-path-to-value.md","skills/paywalls/SKILL.md","skills/paywalls/evals/evals.json","skills/paywalls/references/experiments.md","skills/popups/SKILL.md","skills/popups/evals/evals.json","skills/pricing/SKILL.md","skills/pricing/evals/evals.json","skills/pricing/references/pricing-models.md","skills/pricing/references/pricing-page-teardown.md","skills/pricing/references/research-methods.md","skills/pricing/references/tier-structure.md","skills/product-marketing/SKILL.md","skills/product-marketing/evals/evals.json","skills/product-marketing/references/brand-voice.md","skills/programmatic-seo/SKILL.md","skills/programmatic-seo/evals/evals.json","skills/programmatic-seo/references/implementation-platforms.md","skills/programmatic-seo/references/playbooks.md","skills/prospecting/SKILL.md","skills/prospecting/evals/evals.json","skills/prospecting/references/account-research.md","skills/prospecting/references/b2b-prospecting.md","skills/prospecting/references/compliance.md","skills/prospecting/references/data-sources.md","skills/prospecting/references/demand-signals.md","skills/prospecting/references/enrichment-playbook.md","skills/prospecting/references/local-prospecting.md","skills/prospecting/references/saas-prospecting.md","skills/prospecting/references/signal-plays.md","skills/prospecting/references/sourcing-and-enrichment.md","skills/public-relations/SKILL.md","skills/public-relations/evals/evals.json","skills/public-relations/references/crisis-communications.md","skills/public-relations/references/journalist-pitching.md","skills/public-relations/references/media-outlets.md","skills/public-relations/references/newsjacking.md","skills/public-relations/references/podcast-guest-prep.md","skills/public-relations/references/press-platforms.md","skills/public-relations/references/statement-templates.md","skills/public-relations/references/story-angles.md","skills/referrals/SKILL.md","skills/referrals/evals/evals.json","skills/referrals/references/affiliate-operations.md","skills/referrals/references/affiliate-programs.md","skills/referrals/references/program-examples.md","skills/referrals/references/review-platform-rules.md","skills/referrals/references/review-program-operations.md","skills/referrals/references/review-programs.md","skills/referrals/references/viral-mechanisms.md","skills/revops/SKILL.md","skills/revops/evals/evals.json","skills/revops/references/automation-playbooks.md","skills/revops/references/inbound-routing.md","skills/revops/references/lifecycle-definitions.md","skills/revops/references/routing-rules.md","skills/revops/references/scoring-models.md","skills/sales-enablement/SKILL.md","skills/sales-enablement/evals/evals.json","skills/sales-enablement/references/cold-call-scripts.md","skills/sales-enablement/references/deck-frameworks.md","skills/sales-enablement/references/demo-scripts.md","skills/sales-enablement/references/objection-library.md","skills/sales-enablement/references/one-pager-templates.md","skills/sales-enablement/references/win-loss-analysis.md","skills/schema/SKILL.md","skills/schema/evals/evals.json","skills/schema/references/schema-examples.md","skills/seo-audit/SKILL.md","skills/seo-audit/evals/evals.json","skills/seo-audit/references/ai-writing-detection.md","skills/seo-audit/references/international-seo.md","skills/seo-audit/references/local-seo-pages-and-measurement.md","skills/seo-audit/references/local-seo-profile-and-citations.md","skills/seo-audit/references/local-seo.md","skills/seo-audit/references/rankings-push.md","skills/seo-audit/references/title-tags.md","skills/signup/SKILL.md","skills/signup/evals/evals.json","skills/site-architecture/SKILL.md","skills/site-architecture/evals/evals.json","skills/site-architecture/references/mermaid-templates.md","skills/site-architecture/references/navigation-patterns.md","skills/site-architecture/references/platforms-and-migration.md","skills/site-architecture/references/site-type-templates.md","skills/sms/SKILL.md","skills/sms/evals/evals.json","skills/sms/references/compliance.md","skills/sms/references/platforms.md","skills/sms/references/sequence-templates.md","skills/sms/references/whatsapp.md","skills/social/SKILL.md","skills/social/evals/evals.json","skills/social/references/carousel-frameworks.md","skills/social/references/listening-sources-template.md","skills/social/references/listening.md","skills/social/references/platform-limits.md","skills/social/references/platforms.md","skills/social/references/post-templates.md","skills/social/references/reverse-engineering.md","skills/social/references/short-form-video.md","skills/social/references/x-algorithm.md","skills/video/SKILL.md","skills/video/evals/evals.json","skills/video/references/ai-video-prompting.md","skills/video/references/edit-anatomy.md","skills/video/references/product-demo-recording.md","tools/aisa/README.md","tools/aisa/evals/README.md","tools/clis/README.md","tools/composio/README.md"],"skills":{".agents/plugins/marketplace.json":"",".claude-plugin/marketplace.json":"",".claude-plugin/plugin.json":"",".codex-plugin/plugin.json":"","AGENTS.md":"# AGENTS.md\n\nGuidelines for AI agents working in this repository.\n\n## Repository Overview\n\nThis repository contains **Agent Skills** for AI agents following the [Agent Skills specification](https://agentskills.io/specification.md). Skills install to `.agents/skills/` (the cross-agent standard). This repo also serves as a **Claude Code plugin marketplace** via `.claude-plugin/marketplace.json`.\n\n- **Name**: Marketing Skills\n- **GitHub**: [coreyhaines31/marketingskills](https://github.com/coreyhaines31/marketingskills)\n- **Creator**: Corey Haines\n- **License**: MIT\n\n## Repository Structure\n\n```\nmarketingskills/\n├── .claude-plugin/\n│ └── marketplace.json # Claude Code plugin marketplace manifest\n├── skills/ # Agent Skills\n│ └── skill-name/\n│ └── SKILL.md # Required skill file\n├── tools/\n│ ├── clis/ # Zero-dependency Node.js CLI tools (66 tools)\n│ ├── composio/ # Composio integration layer (quick start + toolkit mapping)\n│ ├── integrations/ # API integration guides per tool\n│ └── REGISTRY.md # Tool index with capabilities\n├── CONTRIBUTING.md\n├── LICENSE\n└── README.md\n```\n\n## Build / Lint / Test Commands\n\n**Skills** are content-only (no build step). Verify manually:\n- YAML frontmatter is valid\n- `name` field matches directory name exactly\n- `name` is 1-64 chars, lowercase alphanumeric and hyphens only\n- `description` is 1-1024 characters\n\n**CLI tools** (`tools/clis/*.js`) are zero-dependency Node.js scripts (Node 18+). Verify with:\n```bash\nnode --check tools/clis/.js # Syntax check\nnode tools/clis/.js # Show usage (no args = help)\nnode tools/clis/.js --dry-run # Preview request without sending\nnode --test tests/clis/*.test.cjs # Contract tests (mocked fetch, no live calls)\n```\n\n## Versioning\n\nTwo version layers, with different rules:\n\n**Repo release version** — `.claude-plugin/plugin.json` `version`, `.claude-plugin/marketplace.json` `metadata.version`, and the `VERSIONS.md` changelog headings all share one x.y.z number:\n\n- **x** — repo-wide changes (restructures, spec changes, breaking changes)\n- **y** — new skill(s) added\n- **z** — updates to existing skills\n\nDo not bump y for content added to an existing skill, no matter how substantial — that's a z release (e.g. a new reference file in ad-creative is 2.8.0 → 2.8.1, not 2.9.0).\n\n**Per-skill version** — `metadata.version` in each SKILL.md, mirrored in the `VERSIONS.md` table. Bump on ANY shipped change to that skill: the update check compares `VERSIONS.md` against users' local skill metadata, so an unbumped change is invisible to installed users. Minor for new capability or description triggers, patch for fixes and clarifications.\n\nBump the repo release version in the same PR that ships the change (2.7.0 and 2.8.0 shipped without touching plugin.json/marketplace.json and needed a catch-up later).\n\nCI enforces both layers on PRs. Fetch first so `origin/main` is current, then run it locally before pushing:\n\n```bash\nnode scripts/check-versions.mjs --base origin/main\n```\n\nIt fails if a changed skill didn't bump `metadata.version`, if `VERSIONS.md` disagrees with any SKILL.md, if the repo version didn't bump (or has no `### x.y.z` block), or if a file in `references/` isn't linked from anywhere in its skill. Changes only under a skill's `evals/` don't need a bump, since they don't change what installed users get.\n\n## Agent Skills Specification\n\nSkills follow the [Agent Skills spec](https://agentskills.io/specification.md).\n\n### Required Frontmatter\n\n```yaml\n---\nname: skill-name\ndescription: What this skill does and when to use it. Include trigger phrases.\n---\n```\n\n### Frontmatter Field Constraints\n\n| Field | Required | Constraints |\n|---------------|----------|------------------------------------------------------------------|\n| `name` | Yes | 1-64 chars, lowercase `a-z`, numbers, hyphens. Must match dir. |\n| `description` | Yes | 1-1024 chars. Describe what it does and when to use it. |\n| `license` | No | License name (default: MIT) |\n| `metadata` | No | Key-value pairs (author, version, etc.) |\n\n### Name Field Rules\n\n- Lowercase letters, numbers, and hyphens only\n- Cannot start or end with hyphen\n- No consecutive hyphens (`--`)\n- Must match parent directory name exactly\n\n**Valid**: `cro`, `emails`, `ab-testing`\n**Invalid**: `Page-CRO`, `-page`, `page--cro`\n\n### Optional Skill Directories\n\n```\nskills/skill-name/\n├── SKILL.md # Required - main instructions (<500 lines)\n├── references/ # Optional - detailed docs loaded on demand\n├── scripts/ # Optional - executable code\n└── assets/ # Optional - templates, data files\n```\n\n## Writing Style Guidelines\n\n### Structure\n\n- Keep `SKILL.md` under 500 lines (move details to `references/`)\n- Use H2 (`##`) for main sections, H3 (`###`) for subsections\n- Use bullet points and numbered lists liberally\n- Short paragraphs (2-4 sentences max)\n\n### Tone\n\n- Direct and instructional\n- Second person (\"You are a conversion rate optimization expert\")\n- Professional but approachable\n\n### Formatting\n\n- Bold (`**text**`) for key terms\n- Code blocks for examples and templates\n- Tables for reference data\n- No excessive emojis\n\n### Clarity Principles\n\n- Clarity over cleverness\n- Specific over vague\n- Active voice over passive\n- One idea per section\n\n### Description Field Best Practices\n\nThe `description` is critical for skill discovery. Include:\n1. What the skill does\n2. When to use it (trigger phrases)\n3. Related skills for scope boundaries\n\n```yaml\ndescription: When the user wants to optimize conversions on any marketing page. Use when the user says \"CRO,\" \"conversion rate optimization,\" \"this page isn't converting.\" For signup flows, see signup.\n```\n\n## Claude Code Plugin\n\nThis repo also serves as a plugin marketplace. The manifest at `.claude-plugin/marketplace.json` lists all skills for installation via:\n\n```bash\n/plugin marketplace add coreyhaines31/marketingskills\n/plugin install marketing-skills\n```\n\nSee [Claude Code plugins documentation](https://code.claude.com/docs/en/plugins.md) for details.\n\n## Git Workflow\n\n### Branch Naming\n\n- New skills: `feature/skill-name`\n- Improvements: `fix/skill-name-description`\n- Documentation: `docs/description`\n\n### Commit Messages\n\nFollow the [Conventional Commits](https://www.conventionalcommits.org/) specification:\n\n- `feat: add skill-name skill`\n- `fix: improve clarity in cro`\n- `docs: update README`\n\n### Pull Request Checklist\n\n- [ ] `name` matches directory name exactly\n- [ ] `name` follows naming rules (lowercase, hyphens, no `--`)\n- [ ] `description` is 1-1024 chars with trigger phrases\n- [ ] `SKILL.md` is under 500 lines\n- [ ] No sensitive data or credentials\n\n## Tool Integrations\n\nThis repository includes a tools registry for agent-compatible marketing tools.\n\n- **Tool discovery**: Read `tools/REGISTRY.md` to see available tools and their capabilities\n- **Connecting accounts**: For a new local setup, follow `tools/SETUP.md`: check the environment, configure only chosen tools, keep secrets local, and verify a read before proposing broader work\n- **Integration details**: See `tools/integrations/{tool}.md` for API endpoints, auth, and common operations\n- **MCP-enabled tools** (checked 2026-10-15; the registry is the source of truth): ga4, stripe, mailchimp, google-ads, meta-ads, resend, zapier, zoominfo, clay, supermetrics, coupler, outreach, crossbeam, introw, exa, apollo, hubspot, attio, close, instantly, lemlist, heyreach, hunter, truelist, fullenrich, leadmagic, rowbound, theirstack, apify, firecrawl, browserbase, calendly, composio\n- **Composio** (integration layer): Adds MCP access to OAuth-heavy tools without native MCP servers (Salesforce, LinkedIn Ads, Google Sheets, Slack, etc.). See `tools/integrations/composio.md`\n\n### Registry Structure\n\n```\ntools/\n├── REGISTRY.md # Index of all tools with capabilities\n└── integrations/ # Detailed integration guides\n ├── ga4.md\n ├── stripe.md\n ├── rewardful.md\n └── ...\n```\n\n### When to Use Tools\n\nSkills reference relevant tools for implementation. For example:\n- `referrals` skill → rewardful, tolt, dub-co, mention-me guides\n- `analytics` skill → ga4, mixpanel, segment guides\n- `emails` skill → customer-io, mailchimp, resend guides\n- `ads` skill → google-ads, meta-ads, linkedin-ads guides\n\nFor tools without native MCP servers (Salesforce, LinkedIn Ads, Google Sheets, Slack, Notion), Composio provides MCP access via a single server. See `tools/integrations/composio.md` for setup and `tools/composio/marketing-tools.md` for the full toolkit mapping.\n\n## Checking for Updates\n\nWhen using any skill from this repository:\n\n1. **Once per session**, on first skill use, check for updates:\n - Fetch `VERSIONS.md` from GitHub: https://raw.githubusercontent.com/coreyhaines31/marketingskills/main/VERSIONS.md\n - Compare versions against local skill files\n\n2. **Only prompt if meaningful**:\n - 2 or more skills have updates, OR\n - Any skill has a major version bump (e.g., 1.x to 2.x)\n\n3. **Non-blocking notification** at end of response:\n ```\n ---\n Skills update available: X marketing skills have updates.\n Say \"update skills\" to update automatically, or run `git pull` in your marketingskills folder.\n ```\n\n4. **If user says \"update skills\"**:\n - Run `git pull` in the marketingskills directory\n - Confirm what was updated\n\n## Skill Categories\n\nSee `README.md` for the current list of skills organized by category. When adding new skills, follow the naming patterns of existing skills in that category.\n\n## Claude Code-Specific Enhancements\n\nThese patterns are **Claude Code only** and must not be added to `SKILL.md` files directly, as skills are designed to be cross-agent compatible (Codex, Cursor, Windsurf, etc.). Apply them locally in your own project's `.claude/skills/` overrides instead.\n\n### Dynamic content injection with `!`command``\n\nClaude Code supports embedding shell commands in SKILL.md using `` !`command` `` syntax. When the skill is invoked, Claude Code runs the command and injects the output inline — the model sees the result, not the instruction.\n\n**Most useful application: auto-inject the product marketing context file**\n\nInstead of every skill telling the agent \"go check if `.agents/product-marketing.md` exists and read it,\" you can inject it automatically:\n\n```markdown\nProduct context: !`cat .agents/product-marketing.md 2>/dev/null || echo \"No product context file found — ask the user about their product before proceeding.\"`\n```\n\nPlace this at the top of a skill's body (after frontmatter) to make context available immediately without any file-reading step.\n\n**Other useful injections:**\n\n```markdown\n# Inject today's date for recency-sensitive skills\nToday's date: !`date +%Y-%m-%d`\n\n# Inject current git branch (useful for workflow skills)\nCurrent branch: !`git branch --show-current 2>/dev/null`\n\n# Inject recent commits for context\nRecent commits: !`git log --oneline -5 2>/dev/null`\n```\n\n**Why this is Claude Code-only**: Other agents that load skills will see the literal `` !`command` `` string rather than executing it, which would appear as garbled instructions. Keep cross-agent skill files free of this syntax.\n","CLAUDE.md":"AGENTS.md","skills/ab-testing/SKILL.md":"---\nname: ab-testing\ndescription: When the user wants to plan, design, or implement an A/B test or experiment, or build a growth experimentation program. Also use when the user mentions \"A/B test,\" \"split test,\" \"experiment,\" \"test this change,\" \"variant copy,\" \"multivariate test,\" \"hypothesis,\" \"should I test this,\" \"which version is better,\" \"test two versions,\" \"statistical significance,\" \"how long should I run this test,\" \"growth experiments,\" \"experiment velocity,\" \"experiment backlog,\" \"ICE score,\" \"experimentation program,\" or \"experiment playbook.\" Use this whenever someone is comparing two approaches and wants to measure which performs better, or when they want to build a systematic experimentation practice. For tracking implementation, see analytics. For page-level conversion optimization, see cro.\nmetadata:\n version: 2.0.2\n---\n\n# A/B Test Setup\n\nYou are an expert in experimentation and A/B testing. Your goal is to help design tests that produce statistically valid, actionable results.\n\n## Initial Assessment\n\n**Check for product marketing context first:**\nIf `.agents/product-marketing.md` exists (or `.claude/product-marketing.md`, or the legacy `product-marketing-context.md` filename, in older setups), read it before asking questions. Use that context and only ask for information not already covered or specific to this task.\n\nBefore designing a test, understand:\n\n1. **Test Context** - What are you trying to improve? What change are you considering?\n2. **Current State** - Baseline conversion rate? Current traffic volume?\n3. **Constraints** - Technical complexity? Timeline? Tools available?\n\n---\n\n## Core Principles\n\n### 1. Start with a Hypothesis\n- Not just \"let's see what happens\"\n- Specific prediction of outcome\n- Based on reasoning or data\n\n### 2. Test One Thing\n- Single variable per test\n- Otherwise you don't know what worked\n\n### 3. Statistical Rigor\n- Pre-determine sample size\n- Don't peek and stop early\n- Commit to the methodology\n\n### 4. Measure What Matters\n- Primary metric tied to business value\n- Secondary metrics for context\n- Guardrail metrics to prevent harm\n\n---\n\n## Hypothesis Framework\n\n### Structure\n\n```\nBecause [observation/data],\nwe believe [change]\nwill cause [expected outcome]\nfor [audience].\nWe'll know this is true when [metrics].\n```\n\n### Example\n\n**Weak**: \"Changing the button color might increase clicks.\"\n\n**Strong**: \"Because users report difficulty finding the CTA (per heatmaps and feedback), we believe making the button larger and using contrasting color will increase CTA clicks by 15%+ for new visitors. We'll measure click-through rate from page view to signup start.\"\n\n---\n\n## Test Types\n\n| Type | Description | Traffic Needed |\n|------|-------------|----------------|\n| A/B | Two versions, single change | Moderate |\n| A/B/n | Multiple variants | Higher |\n| MVT | Multiple changes in combinations | Very high |\n| Split URL | Different URLs for variants | Moderate |\n\n---\n\n## Sample Size\n\n### Quick Reference\n\n| Baseline | 10% Lift | 20% Lift | 50% Lift |\n|----------|----------|----------|----------|\n| 1% | 150k/variant | 39k/variant | 6k/variant |\n| 3% | 47k/variant | 12k/variant | 2k/variant |\n| 5% | 27k/variant | 7k/variant | 1.2k/variant |\n| 10% | 12k/variant | 3k/variant | 550/variant |\n\n**Calculators:**\n- [Evan Miller's](https://www.evanmiller.org/ab-testing/sample-size.html)\n- [Optimizely's](https://www.optimizely.com/sample-size-calculator/)\n\n**For detailed sample size tables and duration calculations**: See [references/sample-size-guide.md](references/sample-size-guide.md)\n\n---\n\n## Metrics Selection\n\n### Primary Metric\n- Single metric that matters most\n- Directly tied to hypothesis\n- What you'll use to call the test\n\n### Secondary Metrics\n- Support primary metric interpretation\n- Explain why/how the change worked\n\n### Guardrail Metrics\n- Things that shouldn't get worse\n- Stop test if significantly negative\n\n### Example: Pricing Page Test\n- **Primary**: Plan selection rate\n- **Secondary**: Time on page, plan distribution\n- **Guardrail**: Support tickets, refund rate\n\n---\n\n## Designing Variants\n\n### What to Vary\n\n| Category | Examples |\n|----------|----------|\n| Headlines/Copy | Message angle, value prop, specificity, tone |\n| Visual Design | Layout, color, images, hierarchy |\n| CTA | Button copy, size, placement, number |\n| Content | Information included, order, amount, social proof |\n\n### Best Practices\n- Single, meaningful change\n- Bold enough to make a difference\n- True to the hypothesis\n\n---\n\n## Traffic Allocation\n\n| Approach | Split | When to Use |\n|----------|-------|-------------|\n| Standard | 50/50 | Default for A/B |\n| Conservative | 90/10, 80/20 | Limit risk of bad variant |\n| Ramping | Start small, increase | Technical risk mitigation |\n\n**Considerations:**\n- Consistency: Users see same variant on return\n- Balanced exposure across time of day/week\n\n---\n\n## Implementation\n\n### Client-Side\n- JavaScript modifies page after load\n- Quick to implement, can cause flicker\n- Tools: PostHog, Optimizely, VWO\n\n### Server-Side\n- Variant determined before render\n- No flicker, requires dev work\n- Tools: PostHog, LaunchDarkly, Split\n\n---\n\n## Running the Test\n\n### Pre-Launch Checklist\n- [ ] Hypothesis documented\n- [ ] Primary metric defined\n- [ ] Sample size calculated\n- [ ] Variants implemented correctly\n- [ ] Tracking verified\n- [ ] QA completed on all variants\n\n### During the Test\n\n**DO:**\n- Monitor for technical issues\n- Check segment quality\n- Document external factors\n\n**Avoid:**\n- Peek at results and stop early\n- Make changes to variants\n- Add traffic from new sources\n\n### The Peeking Problem\nLooking at results before reaching sample size and stopping early leads to false positives and wrong decisions. Pre-commit to sample size and trust the process.\n\n---\n\n## Analyzing Results\n\n### Statistical Significance\n- For a pre-specified fixed-horizon test, compare its p-value with the chosen significance level (commonly α = 0.05).\n- A p-value describes how incompatible the data are with the null model: the probability, **assuming that model and its assumptions**, of a result at least as extreme as the observed one. It is not the probability that the result is random, that the null is true, or that a variant will win again.\n- A 95% confidence interval comes from a procedure with 95% coverage over repeated samples under its assumptions. It is not a 95% posterior probability that this particular interval contains the true effect. Report the interval and effect size alongside the p-value.\n- Use the planned analysis and stopping rule. Sequential tests and Bayesian analyses have their own interpretation; do not translate every platform's “confidence” score into a fixed-horizon p-value.\n\nThese distinctions follow the [American Statistical Association's p-value statement](https://www.amstat.org/asa/files/pdfs/p-valuestatement.pdf). Statistical significance alone does not establish business value or justify shipping a variant.\n\n### Analysis Checklist\n\n1. **Reach sample size?** If not, result is preliminary\n2. **Statistically significant?** Check confidence intervals\n3. **Effect size meaningful?** Compare to MDE, project impact\n4. **Secondary metrics consistent?** Support the primary?\n5. **Guardrail concerns?** Anything get worse?\n6. **Segment differences?** Mobile vs. desktop? New vs. returning?\n\n### Interpreting Results\n\n| Result | Conclusion |\n|--------|------------|\n| Significant winner | Check effect size, uncertainty and guardrails before implementing |\n| Significant loser | Keep control, learn why |\n| No significant difference | Report the effect interval: it may still allow meaningful benefit and harm, or may exclude effects worth pursuing. Follow the planned stopping rule; do not extend a finished fixed-horizon test until it becomes significant. Plan a new test if needed. |\n| Mixed signals | Check data quality and planned metrics; label unplanned segment findings exploratory and confirm them in a new test |\n\n---\n\n## Documentation\n\nDocument every test with:\n- Hypothesis\n- Variants (with screenshots)\n- Results (sample, metrics, significance)\n- Decision and learnings\n\n**For templates**: See [references/test-templates.md](references/test-templates.md)\n\n---\n\n## Growth Experimentation Program\n\nIndividual tests are valuable. A continuous experimentation program is a compounding asset. This section covers how to run experiments as an ongoing growth engine, not just one-off tests.\n\n### The Experiment Loop\n\n```\n1. Generate hypotheses (from data, research, competitors, customer feedback)\n2. Prioritize with ICE scoring\n3. Design and run the test\n4. Analyze results with statistical rigor\n5. Promote winners to a playbook\n6. Generate new hypotheses from learnings\n→ Repeat\n```\n\n### Hypothesis Generation\n\nFeed your experiment backlog from multiple sources:\n\n| Source | What to Look For |\n|--------|-----------------|\n| Analytics | Drop-off points, low-converting pages, underperforming segments |\n| Customer research | Pain points, confusion, unmet expectations |\n| Competitor analysis | Features, messaging, or UX patterns they use that you don't |\n| Support tickets | Recurring questions or complaints about conversion flows |\n| Heatmaps/recordings | Where users hesitate, rage-click, or abandon |\n| Past experiments | \"Significant loser\" tests often reveal new angles to try |\n\n### ICE Prioritization\n\nScore each hypothesis 1-10 on three dimensions:\n\n| Dimension | Question |\n|-----------|----------|\n| **Impact** | If this works, how much will it move the primary metric? |\n| **Confidence** | How sure are we this will work? (Based on data, not gut.) |\n| **Ease** | How fast and cheap can we ship and measure this? |\n\n**ICE Score** = (Impact + Confidence + Ease) / 3\n\nRun highest-scoring experiments first. Re-score monthly as context changes.\n\n### Experiment Velocity\n\nTrack your experimentation rate as a leading indicator of growth:\n\n| Metric | Target |\n|--------|--------|\n| Experiments launched per month | 4-8 for most teams |\n| Win rate | 20-30% is common for mature programs (sustained higher rates may indicate conservative hypotheses) |\n| Average test duration | 2-4 weeks |\n| Backlog depth | 20+ hypotheses queued |\n| Cumulative lift | Compound gains from all winners |\n\n### The Experiment Playbook\n\nWhen a test wins, don't just implement it — document the pattern:\n\n```\n## [Experiment Name]\n**Date**: [date]\n**Hypothesis**: [the hypothesis]\n**Sample size**: [n per variant]\n**Result**: [winner/loser/inconclusive] — [primary metric] changed by [X%] (95% CI: [range], p=[value])\n**Guardrails**: [any guardrail metrics and their outcomes]\n**Segment deltas**: [notable differences by device, segment, or cohort]\n**Why it worked/failed**: [analysis]\n**Pattern**: [the reusable insight — e.g., \"social proof near pricing CTAs increases plan selection\"]\n**Apply to**: [other pages/flows where this pattern might work]\n**Status**: [implemented / parked / needs follow-up test]\n```\n\nOver time, your playbook becomes a library of proven growth patterns specific to your product and audience.\n\n### Experiment Cadence\n\n**Weekly (30 min)**: Review running experiments for technical issues and guardrail metrics. Don't call winners early — but do stop tests where guardrails are significantly negative.\n\n**Bi-weekly**: Conclude completed experiments. Analyze results, update playbook, launch next experiment from backlog.\n\n**Monthly (1 hour)**: Review experiment velocity, win rate, cumulative lift. Replenish hypothesis backlog. Re-prioritize with ICE.\n\n**Quarterly**: Audit the playbook. Which patterns have been applied broadly? Which winning patterns haven't been scaled yet? What areas of the funnel are under-tested?\n\n---\n\n## Common Mistakes\n\n### Test Design\n- Testing too small a change (undetectable)\n- Testing too many things (can't isolate)\n- No clear hypothesis\n\n### Execution\n- Stopping early\n- Changing things mid-test\n- Not checking implementation\n\n### Analysis\n- Ignoring confidence intervals\n- Cherry-picking segments\n- Over-interpreting inconclusive results\n\n---\n\n## Task-Specific Questions\n\n1. What's your current conversion rate?\n2. How much traffic does this page get?\n3. What change are you considering and why?\n4. What's the smallest improvement worth detecting?\n5. What tools do you have for testing?\n6. Have you tested this area before?\n\n---\n\n## Related Skills\n\n- **cro**: For generating test ideas based on CRO principles\n- **analytics**: For setting up test measurement\n- **copywriting**: For creating variant copy\n","skills/ab-testing/evals/evals.json":"{\n \"skill_name\": \"ab-testing\",\n \"evals\": [\n {\n \"id\": 1,\n \"prompt\": \"I want to A/B test our homepage headline. We currently say 'The All-in-One Project Management Tool' and want to test something benefit-focused. We get about 15,000 visitors/month and our current signup rate is 3.2%.\",\n \"expected_output\": \"Should check for product-marketing.md first. Should build a proper hypothesis using the framework: 'Because [observation], we believe [change] will cause [outcome], which we'll measure by [metric].' Should identify this as an A/B test (two variants). Should calculate or reference sample size needs based on 15,000 monthly visitors and 3.2% baseline. Should define primary metric (signup rate), secondary metrics, and guardrail metrics. Should warn about the peeking problem and recommend a fixed test duration. Should provide the test plan in the structured output format.\",\n \"assertions\": [\n \"Checks for product-marketing.md\",\n \"Uses the hypothesis framework with observation, belief, outcome, and metric\",\n \"Identifies as A/B test type\",\n \"Addresses sample size calculation based on traffic and baseline rate\",\n \"Defines primary metric (signup rate)\",\n \"Defines secondary and guardrail metrics\",\n \"Warns about the peeking problem\",\n \"Provides structured test plan output\"\n ],\n \"files\": []\n },\n {\n \"id\": 2,\n \"prompt\": \"we want to test like 4 different CTA button colors on our pricing page. is that a good idea?\",\n \"expected_output\": \"Should trigger on casual phrasing. Should identify this as an A/B/n test (multiple variants). Should caution that testing 4 variants requires significantly more traffic than a simple A/B test. Should reference the sample size quick reference showing traffic multipliers for multiple variants. Should question whether button color alone is likely to produce meaningful lift vs testing CTA copy, placement, or surrounding context. Should recommend either reducing to 2 variants or ensuring sufficient traffic. Should still provide hypothesis framework and test setup if proceeding.\",\n \"assertions\": [\n \"Triggers on casual phrasing\",\n \"Identifies as A/B/n test (multiple variants)\",\n \"Cautions about increased traffic needs for 4 variants\",\n \"References sample size requirements\",\n \"Questions whether button color alone is high-impact\",\n \"Suggests alternative higher-impact elements to test\",\n \"Provides hypothesis framework\"\n ],\n \"files\": []\n },\n {\n \"id\": 3,\n \"prompt\": \"Our test has been running for 3 days and Variant B is winning with 95% confidence. Should we call it?\",\n \"expected_output\": \"Should immediately address the peeking problem. Should explain that checking results early inflates false positive rates. Should recommend running for the full pre-calculated duration regardless of early results. Should explain why early significance can be misleading (regression to the mean, day-of-week effects, audience mix shifts). Should provide guidance on when it IS appropriate to stop early (sequential testing methods). Should recommend the pre-test commitment to duration.\",\n \"assertions\": [\n \"Addresses the peeking problem directly\",\n \"Explains why early significance is misleading\",\n \"Recommends running for full pre-calculated duration\",\n \"Mentions day-of-week effects or audience mix shifts\",\n \"Explains false positive rate inflation from peeking\",\n \"Mentions sequential testing as alternative approach\"\n ],\n \"files\": []\n },\n {\n \"id\": 4,\n \"prompt\": \"Help me set up a multivariate test on our landing page. I want to test the headline, hero image, and CTA button simultaneously.\",\n \"expected_output\": \"Should identify this as a Multivariate Test (MVT). Should explain that MVT tests combinations of elements and requires much more traffic than A/B tests. Should calculate or reference traffic needs (combinations multiply: e.g., 2 headlines × 2 images × 2 CTAs = 8 combinations). Should recommend MVT only if traffic supports it, otherwise suggest sequential A/B tests. Should build hypotheses for each element being tested. Should define interaction effects to watch for. Should provide structured test plan.\",\n \"assertions\": [\n \"Identifies as multivariate test (MVT)\",\n \"Explains MVT tests combinations of elements\",\n \"Addresses dramatically higher traffic requirements\",\n \"Calculates number of combinations\",\n \"Suggests sequential A/B tests as alternative if traffic insufficient\",\n \"Builds hypotheses for each element\",\n \"Provides structured test plan\"\n ],\n \"files\": []\n },\n {\n \"id\": 5,\n \"prompt\": \"What metrics should I track for an A/B test on our trial signup page? We're testing a longer form (adds company size and role fields) against the current short form.\",\n \"expected_output\": \"Should apply the metrics selection framework with three tiers: primary, secondary, and guardrail metrics. Primary: form completion rate (the direct conversion metric). Secondary: lead quality metrics (SQL conversion rate, activation rate post-signup). Guardrail: overall signup volume (ensure longer form doesn't tank total signups below acceptable threshold). Should explain the tradeoff between conversion quantity and lead quality. Should note that this test needs longer observation window to measure downstream metrics.\",\n \"assertions\": [\n \"Applies three-tier metric framework (primary, secondary, guardrail)\",\n \"Identifies form completion rate as primary metric\",\n \"Identifies lead quality as secondary metric\",\n \"Defines guardrail metrics to protect against negative outcomes\",\n \"Explains quantity vs quality tradeoff\",\n \"Notes need for longer observation window for downstream metrics\"\n ],\n \"files\": []\n },\n {\n \"id\": 6,\n \"prompt\": \"Can you help me write copy for our new landing page? We want to test it against the current version.\",\n \"expected_output\": \"Should recognize this is primarily a copywriting task, not a test setup task. Should defer to or cross-reference the copywriting skill for writing the actual copy. May help frame the test hypothesis and setup, but should make clear that copywriting is the right skill for creating the page copy itself.\",\n \"assertions\": [\n \"Recognizes this as primarily a copywriting task\",\n \"References or defers to copywriting skill\",\n \"Does not attempt to write full page copy using test setup patterns\",\n \"May offer to help with test hypothesis and setup\"\n ],\n \"files\": []\n },\n {\n \"id\": 7,\n \"prompt\": \"We ran an A/B test on our pricing page for 4 weeks. Control: 2.1% conversion. Variant: 2.4% conversion. 12,000 visitors per variant. Is this statistically significant? Should we ship it?\",\n \"expected_output\": \"Should calculate the result under a declared analysis: 252/12,000 vs 288/12,000 gives approximately p = 0.117 for an uncorrected two-sided pooled-proportion z-test, so it does not meet alpha = 0.05. Should distinguish the 0.3 percentage-point / ~14% relative estimate from evidence for a reliably positive effect. Report uncertainty and practical business impact. Respect the pre-specified duration and stopping rule; do not recommend continuing the completed fixed-horizon test until significant. Unplanned segments are exploratory; a follow-up test needs its own plan.\",\n \"assertions\": [\n \"Uses actual conversion counts and states the analysis assumptions\",\n \"Does not call p ~0.117 significant at alpha = 0.05\",\n \"Distinguishes statistical uncertainty from business value\",\n \"Does not extend a completed fixed-horizon test until significant\",\n \"Labels unplanned segment findings exploratory\"\n ],\n \"files\": []\n },\n {\n \"id\": 8,\n \"prompt\": \"Our fixed-horizon A/B test reached its planned sample today with p = 0.04. Can I tell the CEO there is a 96% chance the change works and just a 4% chance the lift was random? We have a +2% estimate, an interval of +0.1% to +3.9%, and a higher refund rate in the variant.\",\n \"expected_output\": \"Reject both probability interpretations. Explain the p-value conditional on the null model and assumptions, and repeated-sampling interval coverage without asserting a posterior probability. Consider the effect interval, practical impact and refund guardrail before recommending rollout. Do not substitute a numerical significance threshold for the business decision.\",\n \"assertions\": [\n \"Does not interpret 1-p as the probability the variant works\",\n \"Does not interpret p as the probability the result was random\",\n \"Explains the conditional null-model meaning\",\n \"Uses interval, business effect and refund guardrail before rollout\"\n ],\n \"files\": []\n }\n ]\n}\n","skills/ab-testing/references/sample-size-guide.md":"# Sample Size Guide\n\nReference for calculating sample sizes and test duration.\n\n## Contents\n- Sample Size Fundamentals (required inputs, what these mean)\n- Sample Size Quick Reference Tables\n- Duration Calculator (formula, examples, minimum duration rules, maximum duration guidelines)\n- Online Calculators\n- Adjusting for Multiple Variants\n- Common Sample Size Mistakes\n- When Sample Size Requirements Are Too High\n- Sequential Testing\n- Quick Decision Framework\n\n## Sample Size Fundamentals\n\n### Required Inputs\n\n1. **Baseline conversion rate**: Your current rate\n2. **Minimum detectable effect (MDE)**: Smallest change worth detecting\n3. **Statistical significance level**: Usually α = 0.05 (5%)\n4. **Statistical power**: Usually 80% (β = 0.20)\n\n### What These Mean\n\n**Baseline conversion rate**: If your page converts at 5%, that's your baseline.\n\n**MDE (Minimum Detectable Effect)**: The smallest improvement you care about detecting. Set this based on:\n- Business impact (is a 5% lift meaningful?)\n- Implementation cost (worth the effort?)\n- Realistic expectations (what have past tests shown?)\n\n**Significance level (α = 0.05)**: For the planned test and its assumptions, limits the long-run probability of rejecting a true null hypothesis to 5%. It is not the probability that an observed difference is “due to chance.” Repeated unadjusted looks or multiple comparisons do not retain that error rate. See the [ASA's interpretation guidance](https://www.amstat.org/asa/files/pdfs/p-valuestatement.pdf).\n\n**Statistical power (80%)**: Means if there's a real effect of size MDE, you have 80% chance of detecting it.\n\n---\n\n## Sample Size Quick Reference Tables\n\n### Conversion Rate: 1%\n\n| Lift to Detect | Sample per Variant | Total Sample |\n|----------------|-------------------|--------------|\n| 5% (1% → 1.05%) | 1,500,000 | 3,000,000 |\n| 10% (1% → 1.1%) | 380,000 | 760,000 |\n| 20% (1% → 1.2%) | 97,000 | 194,000 |\n| 50% (1% → 1.5%) | 16,000 | 32,000 |\n| 100% (1% → 2%) | 4,200 | 8,400 |\n\n### Conversion Rate: 3%\n\n| Lift to Detect | Sample per Variant | Total Sample |\n|----------------|-------------------|--------------|\n| 5% (3% → 3.15%) | 480,000 | 960,000 |\n| 10% (3% → 3.3%) | 120,000 | 240,000 |\n| 20% (3% → 3.6%) | 31,000 | 62,000 |\n| 50% (3% → 4.5%) | 5,200 | 10,400 |\n| 100% (3% → 6%) | 1,400 | 2,800 |\n\n### Conversion Rate: 5%\n\n| Lift to Detect | Sample per Variant | Total Sample |\n|----------------|-------------------|--------------|\n| 5% (5% → 5.25%) | 280,000 | 560,000 |\n| 10% (5% → 5.5%) | 72,000 | 144,000 |\n| 20% (5% → 6%) | 18,000 | 36,000 |\n| 50% (5% → 7.5%) | 3,100 | 6,200 |\n| 100% (5% → 10%) | 810 | 1,620 |\n\n### Conversion Rate: 10%\n\n| Lift to Detect | Sample per Variant | Total Sample |\n|----------------|-------------------|--------------|\n| 5% (10% → 10.5%) | 130,000 | 260,000 |\n| 10% (10% → 11%) | 34,000 | 68,000 |\n| 20% (10% → 12%) | 8,700 | 17,400 |\n| 50% (10% → 15%) | 1,500 | 3,000 |\n| 100% (10% → 20%) | 400 | 800 |\n\n### Conversion Rate: 20%\n\n| Lift to Detect | Sample per Variant | Total Sample |\n|----------------|-------------------|--------------|\n| 5% (20% → 21%) | 60,000 | 120,000 |\n| 10% (20% → 22%) | 16,000 | 32,000 |\n| 20% (20% → 24%) | 4,000 | 8,000 |\n| 50% (20% → 30%) | 700 | 1,400 |\n| 100% (20% → 40%) | 200 | 400 |\n\n---\n\n## Duration Calculator\n\n### Formula\n\n```\nDuration (days) = (Sample per variant × Number of variants) / (Daily traffic × % exposed)\n\nFor unequal allocation, calculate each arm separately:\nDays per arm = Required sample for that arm / (Daily eligible traffic × Exposure fraction × Arm allocation fraction)\nTotal collection time = longest arm duration\n```\n\nUse fractions for exposure and allocation (50% = 0.5). Count eligible randomized units consistently with the sample-size calculation; repeat pageviews are not additional participants. The balanced formula assumes equal sample requirements and equal allocation.\n\n### Examples\n\n**Scenario 1: High-traffic page**\n- Need: 10,000 per variant (2 variants = 20,000 total)\n- Daily traffic: 5,000 visitors\n- 100% exposed to test\n- Duration: 20,000 / 5,000 = **4 days**\n\n**Scenario 2: Medium-traffic page**\n- Need: 30,000 per variant (60,000 total)\n- Daily traffic: 2,000 visitors\n- 100% exposed\n- Duration: 60,000 / 2,000 = **30 days**\n\n**Scenario 3: Low-traffic with partial exposure**\n- Need: 15,000 per variant (30,000 total)\n- Daily traffic: 500 visitors\n- 50% exposed to test\n- Effective daily: 250\n- Duration: 30,000 / 250 = **120 days** (too long!)\n\n### Minimum Duration Rules\n\nEven with sufficient sample size, run tests for at least:\n- **1 full week**: To capture day-of-week variation\n- **2 business cycles**: If B2B (weekday vs. weekend patterns)\n- **Through paydays**: If e-commerce (beginning/end of month)\n\n### Maximum Duration Guidelines\n\nAvoid running tests longer than 4-8 weeks:\n- Novelty effects wear off\n- External factors intervene\n- Opportunity cost of other tests\n\n---\n\n## Online Calculators\n\n### Recommended Tools\n\n**Evan Miller's Calculator**\nhttps://www.evanmiller.org/ab-testing/sample-size.html\n- Simple interface\n- Bookmark-worthy\n\n**Optimizely's Calculator**\nhttps://www.optimizely.com/sample-size-calculator/\n- Business-friendly language\n- Duration estimates\n\n**AB Test Guide Calculator**\nhttps://www.abtestguide.com/calc/\n- Includes Bayesian option\n- Multiple test types\n\n**VWO Duration Calculator**\nhttps://vwo.com/tools/ab-test-duration-calculator/\n- Duration-focused\n- Good for planning\n\n---\n\n## Adjusting for Multiple Variants\n\nWith more than 2 variants (A/B/n tests), you need more sample:\n\n| Variants | Multiplier |\n|----------|------------|\n| 2 (A/B) | 1x |\n| 3 (A/B/C) | ~1.5x |\n| 4 (A/B/C/D) | ~2x |\n| 5+ | Consider reducing variants |\n\n**Why?** More comparisons increase chance of false positives. You're comparing:\n- A vs B\n- A vs C\n- B vs C (sometimes)\n\nApply Bonferroni correction or use tools that handle this automatically.\n\n---\n\n## Common Sample Size Mistakes\n\n### 1. Underpowered tests\n**Problem**: Not enough sample to detect realistic effects\n**Fix**: Be realistic about MDE, get more traffic, or don't test\n\n### 2. Overpowered tests\n**Problem**: Waiting for sample size when you already have significance\n**Fix**: This is actually fine—you committed to sample size, honor it\n\n### 3. Wrong baseline rate\n**Problem**: Using wrong conversion rate for calculation\n**Fix**: Use the specific metric and page, not site-wide averages\n\n### 4. Ignoring segments\n**Problem**: Calculating for full traffic, then analyzing segments\n**Fix**: If you plan segment analysis, calculate sample for smallest segment\n\n### 5. Testing too many things\n**Problem**: Dividing traffic too many ways\n**Fix**: Prioritize ruthlessly, run fewer concurrent tests\n\n---\n\n## When Sample Size Requirements Are Too High\n\nOptions when you can't get enough traffic:\n\n1. **Increase MDE**: Accept only detecting larger effects (20%+ lift)\n2. **Lower confidence**: Use 90% instead of 95% (risky, document it)\n3. **Reduce variants**: Test only the most promising variant\n4. **Combine traffic**: Test across multiple similar pages\n5. **Test upstream**: Test earlier in funnel where traffic is higher\n6. **Don't test**: Make decision based on qualitative data instead\n7. **Longer test**: Accept longer duration (weeks/months)\n\n---\n\n## Sequential Testing\n\nIf you must check results before reaching sample size:\n\n### What is it?\nStatistical method that adjusts for multiple looks at data.\n\n### When to use\n- High-risk changes\n- Need to stop bad variants early\n- Time-sensitive decisions\n\n### Tools that support it\n- Optimizely (Stats Accelerator)\n- VWO (SmartStats)\n- PostHog (Bayesian approach)\n\n### Tradeoff\n- More flexibility to stop early\n- Slightly larger sample size requirement\n- More complex analysis\n\n---\n\n## Quick Decision Framework\n\n### Can I run this test?\n\n```\nDaily eligible traffic (randomization units): _____\nFraction exposed to the experiment: _____\nNumber of arms (including control): _____\nAllocation fraction for each arm: _____\nBaseline conversion rate: _____\nMDE I care about: _____\n\nSample needed per variant: _____ (from tables above)\nBalanced allocation: Days = Sample per arm × Number of arms / (Daily eligible traffic × Exposure fraction)\nUnequal allocation: Days = max across arms of [Sample required for arm / (Daily eligible traffic × Exposure fraction × Arm allocation)]\nRound collection time up, then apply the minimum-duration rules above.\n\nIf days > 60: Consider alternatives\nIf days > 30: Acceptable for high-impact tests\nIf days < 14: Likely feasible\nIf days < 7: Easy to run, consider running longer anyway\n```\n\n**Duration check (illustrative):** with 2,000 participants required per arm and 1,000 eligible participants/day, two arms at 50/50 with full exposure require four days, not two. At 50% exposure they require eight days. With full exposure and 90/10 allocation, the smaller arm needs 20 days to reach 2,000. These are collection estimates; honor planned business cycles and the analysis stopping rule.\n","skills/ab-testing/references/test-templates.md":"# A/B Test Templates Reference\n\nTemplates for planning, documenting, and analyzing experiments.\n\n## Contents\n- Test Plan Template\n- Results Documentation Template\n- Test Repository Entry Template\n- Quick Test Brief Template\n- Stakeholder Update Template\n- Experiment Prioritization Scorecard\n- Hypothesis Bank Template\n\n## Test Plan Template\n\n```markdown\n# A/B Test: [Name]\n\n## Overview\n- **Owner**: [Name]\n- **Test ID**: [ID in testing tool]\n- **Page/Feature**: [What's being tested]\n- **Planned dates**: [Start] - [End]\n\n## Hypothesis\n\nBecause [observation/data],\nwe believe [change]\nwill cause [expected outcome]\nfor [audience].\nWe'll know this is true when [metrics].\n\n## Test Design\n\n| Element | Details |\n|---------|---------|\n| Test type | A/B / A/B/n / MVT |\n| Duration | X weeks |\n| Sample size | X per variant |\n| Traffic allocation | 50/50 |\n| Tool | [Tool name] |\n| Implementation | Client-side / Server-side |\n\n## Variants\n\n### Control (A)\n[Screenshot]\n- Current experience\n- [Key details about current state]\n\n### Variant (B)\n[Screenshot or mockup]\n- [Specific change #1]\n- [Specific change #2]\n- Rationale: [Why we think this will win]\n\n## Metrics\n\n### Primary\n- **Metric**: [metric name]\n- **Definition**: [how it's calculated]\n- **Current baseline**: [X%]\n- **Minimum detectable effect**: [X%]\n\n### Secondary\n- [Metric 1]: [what it tells us]\n- [Metric 2]: [what it tells us]\n- [Metric 3]: [what it tells us]\n\n### Guardrails\n- [Metric that shouldn't get worse]\n- [Another safety metric]\n\n## Segment Analysis Plan\n- Mobile vs. desktop\n- New vs. returning visitors\n- Traffic source\n- [Other relevant segments]\n\n## Success Criteria\n- Winner: [Primary metric improves by X% with 95% confidence]\n- Loser: [Primary metric decreases significantly]\n- Inconclusive: [What we'll do if no significant result]\n\n## Pre-Launch Checklist\n- [ ] Hypothesis documented and reviewed\n- [ ] Primary metric defined and trackable\n- [ ] Sample size calculated\n- [ ] Test duration estimated\n- [ ] Variants implemented correctly\n- [ ] Tracking verified in all variants\n- [ ] QA completed on all variants\n- [ ] Stakeholders informed\n- [ ] Calendar hold for analysis date\n```\n\n---\n\n## Results Documentation Template\n\n```markdown\n# A/B Test Results: [Name]\n\n## Summary\n| Element | Value |\n|---------|-------|\n| Test ID | [ID] |\n| Dates | [Start] - [End] |\n| Duration | X days |\n| Result | Winner / Loser / Inconclusive |\n| Decision | [What we're doing] |\n\n## Hypothesis (Reminder)\n[Copy from test plan]\n\n## Results\n\n### Sample Size\n| Variant | Target | Actual | % of target |\n|---------|--------|--------|-------------|\n| Control | X | Y | Z% |\n| Variant | X | Y | Z% |\n\n### Primary Metric: [Metric Name]\n| Variant | Value | 95% CI | vs. Control |\n|---------|-------|--------|-------------|\n| Control | X% | [X%, Y%] | — |\n| Variant | X% | [X%, Y%] | +X% |\n\n**Statistical significance**: p = X.XX (95% = sig / not sig)\n**Practical significance**: [Is this lift meaningful for the business?]\n\n### Secondary Metrics\n\n| Metric | Control | Variant | Change | Significant? |\n|--------|---------|---------|--------|--------------|\n| [Metric 1] | X | Y | +Z% | Yes/No |\n| [Metric 2] | X | Y | +Z% | Yes/No |\n\n### Guardrail Metrics\n\n| Metric | Control | Variant | Change | Concern? |\n|--------|---------|---------|--------|----------|\n| [Metric 1] | X | Y | +Z% | Yes/No |\n\n### Segment Analysis\n\n**Mobile vs. Desktop**\n| Segment | Control | Variant | Lift |\n|---------|---------|---------|------|\n| Mobile | X% | Y% | +Z% |\n| Desktop | X% | Y% | +Z% |\n\n**New vs. Returning**\n| Segment | Control | Variant | Lift |\n|---------|---------|---------|------|\n| New | X% | Y% | +Z% |\n| Returning | X% | Y% | +Z% |\n\n## Interpretation\n\n### What happened?\n[Explanation of results in plain language]\n\n### Why do we think this happened?\n[Analysis and reasoning]\n\n### Caveats\n[Any limitations, external factors, or concerns]\n\n## Decision\n\n**Winner**: [Control / Variant]\n\n**Action**: [Implement variant / Keep control / Re-test]\n\n**Timeline**: [When changes will be implemented]\n\n## Learnings\n\n### What we learned\n- [Key insight 1]\n- [Key insight 2]\n\n### What to test next\n- [Follow-up test idea 1]\n- [Follow-up test idea 2]\n\n### Impact\n- **Projected lift**: [X% improvement in Y metric]\n- **Business impact**: [Revenue, conversions, etc.]\n```\n\n---\n\n## Test Repository Entry Template\n\nFor tracking all tests in a central location:\n\n```markdown\n| Test ID | Name | Page | Dates | Primary Metric | Result | Lift | Link |\n|---------|------|------|-------|----------------|--------|------|------|\n| 001 | Hero headline test | Homepage | 1/1-1/15 | CTR | Winner | +12% | [Link] |\n| 002 | Pricing table layout | Pricing | 1/10-1/31 | Plan selection | Loser | -5% | [Link] |\n| 003 | Signup form fields | Signup | 2/1-2/14 | Completion | Inconclusive | +2% | [Link] |\n```\n\n---\n\n## Quick Test Brief Template\n\nFor simple tests that don't need full documentation:\n\n```markdown\n## [Test Name]\n\n**What**: [One sentence description]\n**Why**: [One sentence hypothesis]\n**Metric**: [Primary metric]\n**Duration**: [X weeks]\n**Result**: [TBD / Winner / Loser / Inconclusive]\n**Learnings**: [Key takeaway]\n```\n\n---\n\n## Stakeholder Update Template\n\n```markdown\n## A/B Test Update: [Name]\n\n**Status**: Running / Complete\n**Days remaining**: X (or complete)\n**Current sample**: X% of target\n\n### Preliminary observations\n[What we're seeing - without making decisions yet]\n\n### Next steps\n[What happens next]\n\n### Timeline\n- [Date]: Analysis complete\n- [Date]: Decision and recommendation\n- [Date]: Implementation (if winner)\n```\n\n---\n\n## Experiment Prioritization Scorecard\n\nFor deciding which tests to run:\n\n| Factor | Weight | Test A | Test B | Test C |\n|--------|--------|--------|--------|--------|\n| Potential impact | 30% | | | |\n| Confidence in hypothesis | 25% | | | |\n| Ease of implementation | 20% | | | |\n| Risk if wrong | 15% | | | |\n| Strategic alignment | 10% | | | |\n| **Total** | | | | |\n\nScoring: 1-5 (5 = best)\n\n---\n\n## Hypothesis Bank Template\n\nFor collecting test ideas:\n\n```markdown\n| ID | Page/Area | Observation | Hypothesis | Potential Impact | Status |\n|----|-----------|-------------|------------|------------------|--------|\n| H1 | Homepage | Low scroll depth | Shorter hero will increase scroll | High | Testing |\n| H2 | Pricing | Users compare plans | Comparison table will help | Medium | Backlog |\n| H3 | Signup | Drop-off at email | Social login will increase completion | Medium | Backlog |\n```\n","skills/ad-creative/SKILL.md":"---\nname: ad-creative\ndescription: \"When the user wants to generate, iterate, or scale ad creative — headlines, descriptions, primary text, or full ad variations — for any paid advertising platform. Also use when the user mentions 'ad copy variations,' 'ad creative,' 'generate headlines,' 'RSA headlines,' 'bulk ad copy,' 'creative testing,' 'write me some ads,' 'Facebook ad copy,' 'Google ad headlines,' 'LinkedIn ad text,' 'static ads,' 'ad templates,' 'iMessage ad,' 'ChatGPT ad,' 'creative strategy,' 'creative roadmap,' 'creative retro,' 'hook writing,' 'creative review page,' 'motion video ad,' 'faceless video ad,' 'UGC ad,' 'greenscreen ad,' 'TikTok/Reels ad format,' 'Meta ad format tier list,' or 'creative format taxonomy.' Use it to produce or iterate ad copy at scale. Copy avoids AI tells like 'it's not X, it's Y' reveals. For campaign strategy and targeting, see ads. For landing page copy, see copywriting. Test hypotheses must be falsifiable, naming the metric, the comparison, and the expected direction.\"\nmetadata:\n version: 2.12.1\n---\n\n# Ad Creative\n\nYou are an expert performance creative strategist. Your goal is to generate high-performing ad creative at scale — headlines, descriptions, and primary text that drive clicks and conversions — and iterate based on real performance data.\n\n## Before Starting\n\n**Check for product marketing context first:**\nIf `.agents/product-marketing.md` exists (or `.claude/product-marketing.md`, or the legacy `product-marketing-context.md` filename, in older setups), read it before asking questions. Use that context and only ask for information not already covered or specific to this task.\n\nCheck its **Learnings** section before drafting: angles, offers, and formats already tested there shouldn't be re-proposed as new ideas, and active lessons apply unless the user says otherwise.\n\nGather this context (ask if not provided):\n\n### 1. Platform & Format\n- What platform? (Google Ads, Meta, LinkedIn, TikTok, Twitter/X)\n- What ad format? (Search RSAs, display, social feed, stories, video)\n- Are there existing ads to iterate on, or starting from scratch?\n\n### 2. Product & Offer\n- What are you promoting? (Product, feature, free trial, demo, lead magnet)\n- What's the core value proposition?\n- What makes this different from competitors?\n\n### 3. Audience & Intent\n- Who is the target audience?\n- What stage of awareness? (Problem-aware, solution-aware, product-aware)\n- What pain points or desires drive them?\n\n### 4. Performance Data (if iterating)\n- What creative is currently running?\n- Which headlines/descriptions are performing best? (CTR, conversion rate, ROAS)\n- Which are underperforming?\n- What angles or themes have been tested?\n\n### 5. Constraints\n- Brand voice guidelines or words to avoid?\n- Compliance requirements? (Industry regulations, platform policies)\n- Any mandatory elements? (Brand name, trademark symbols, disclaimers)\n\n---\n\n## How This Skill Works\n\nThis skill supports four modes:\n\n### Mode 1: Generate from Scratch\nWhen starting fresh, you generate a full set of ad creative based on product context, audience insights, and platform best practices.\n\n### Mode 2: Iterate from Performance Data\nWhen the user provides performance data (CSV, paste, or API output), you analyze what's working, identify patterns in top performers, and generate new variations that build on winning themes while exploring new angles.\n\nThe core loop:\n\n```\nPull performance data → Identify winning patterns → Generate new variations → Validate specs → Deliver\n```\n\n### Mode 3: Scaled Static Batches (Grounded)\nFor recurring static ad production at volume (e.g., 50 concepts per batch), work from a **grounded inputs corpus** and the [static ad template library](references/static-ad-templates.md). Every concept must trace to real source material — see \"Grounded Inputs\" below. To run this on a daily or weekly cadence, see the daily-creative-drop loop in **marketing-loops**. To present a batch for client or stakeholder approval, produce a [creative review page](references/creative-review-page.md).\n\nFor **text-led static headlines**, select treatments by the reader's awareness stage and deliver a mobile-readable visual spec using [references/awareness-headlines.md](references/awareness-headlines.md). It complements the static templates and hook system; numerical claims still need their own evidence.\n\n### Mode 4: Creative Strategy Loop\nFor deciding **which ads are worth making before making them**: synthesize three signal sources (account performance, customer language, external organic) into evidence-ranked concepts, branch the creative mix on account state (exploration vs. scaling), maintain a capacity-checked roadmap with production tiers, and run a monthly retro that feeds the next slate. The full system lives in [references/creative-roadmap.md](references/creative-roadmap.md); for hook generation and funnel-stage diagnosis inside any mode, load [references/hook-system.md](references/hook-system.md).\n\n---\n\n## Grounded Inputs\n\nMost AI ad generation fails on input grounding, not output quality: ungrounded generation produces plausible-sounding ads based on training data, not on what converts for this brand. For scaled production (Mode 3), maintain a durable inputs corpus:\n\n```\ninputs/\n winning-ads/ 10-20 screenshots of the highest-performing ads from the last 90 days\n reviews/ 50-100 customer reviews (Trustpilot, G2, Amazon, App Store) as .md/.txt\n comments/ Top comments from existing ad campaigns — objections, unprompted praise, customer-raised angles\nbrand/ Brand voice doc, hex codes, logo, product/screenshot assets\noutputs/ Dated batch folders (outputs/YYYY-MM-DD/)\n```\n\n**Why each input matters:**\n- **Winning ads** carry the hooks, structures, and angles already proven for this brand\n- **Reviews** carry the exact language buyers use for pain, transformation, and unexpected benefits — pull copy from them verbatim rather than paraphrasing\n- **Ad comments** are the most-skipped and highest-value input: objections (\"but does it work for X?\") become FAQ Card ads, and unprompted praise surfaces angles you didn't write\n\n**Grounding rules:**\n- Every concept cites its source (which review, winning ad, or comment it traces to)\n- No invented claims, stats, or testimonials — ever\n- If `inputs/winning-ads/` or `inputs/reviews/` is empty, stop and ask the user to populate it before generating. Do not generate ungrounded concepts as a fallback.\n- Inputs decay: refresh `inputs/winning-ads/` as new ads scale; refresh `inputs/reviews/` and `inputs/comments/` monthly\n\n---\n\n## Platform Specs\n\nPlatforms reject or truncate creative that exceeds these limits, so verify every piece of copy fits before delivering.\n\n### Google Ads (Responsive Search Ads)\n\n| Element | Limit | Quantity |\n|---------|-------|----------|\n| Headline | 30 characters | Up to 15 |\n| Description | 90 characters | Up to 4 |\n| Display URL path | 15 characters each | 2 paths |\n\n**RSA rules:**\n- Headlines must make sense independently and in any combination\n- Pin headlines to positions only when necessary (reduces optimization)\n- Include at least one keyword-focused headline\n- Include at least one benefit-focused headline\n- Include at least one CTA headline\n\n### Meta Ads (Facebook/Instagram)\n\n| Element | Limit | Notes |\n|---------|-------|-------|\n| Primary text | 125 chars visible (up to 2,200) | Front-load the hook |\n| Headline | 40 characters recommended | Below the image |\n| Description | 30 characters recommended | Below headline |\n| URL display link | 40 characters | Optional |\n\n### LinkedIn Ads\n\n| Element | Limit | Notes |\n|---------|-------|-------|\n| Intro text | 150 chars recommended (600 max) | Above the image |\n| Headline | 70 chars recommended (200 max) | Below the image |\n| Description | 100 chars recommended (300 max) | Appears in some placements |\n\n### TikTok Ads\n\n| Element | Limit | Notes |\n|---------|-------|-------|\n| Ad text | 80 chars recommended (100 max) | Above the video |\n| Display name | 40 characters | Brand name |\n\n### Twitter/X Ads\n\n| Element | Limit | Notes |\n|---------|-------|-------|\n| Tweet text | 280 characters | The ad copy |\n| Headline | 70 characters | Card headline |\n| Description | 200 characters | Card description |\n\nFor detailed specs and format variations, see [references/platform-specs.md](references/platform-specs.md).\n\n---\n\n## Generating Ad Visuals\n\n**To decide *which format to make next*** (before briefing any specific ad), consult the Meta creative format taxonomy in [references/meta-creative-formats.md](references/meta-creative-formats.md) — a prioritized S→F catalog of ~51 formats ranked by one question: is it a *unicorn scaler* that punctures cold net-new audiences, or a *supporting cast* member that only converts mid-funnel? Leads with the persona-based Andromeda context (why creator-fronted formats top the list), S-tier callouts (founder content, partnership ads, VSL), the A-tier bench, and explicit F-tier de-prioritization (press, podcast, notes-app fake-native). Use it to pick a format and build a portfolio; the how-to-build detail lives in the static/video references below. For the account-level kill/keep/scale math once ads are live, cross-reference the `ads` skill's the **ads** skill's Meta decision system reference.\n\n**For static ad structure**, use the template library in [references/static-ad-templates.md](references/static-ad-templates.md) — layout frameworks (Us vs. Them, Stat Callout, Review Card, Before/After, Founder Message, FAQ Card, Grid Static, Callout, and more) with copy slots, DTC and SaaS examples, and per-concept output format. Each template carries a **tier (S–F)** and **funnel role** (unicorn cold-scaler vs. mid-funnel supporting cast) so you reach for the right one first. Cycle through templates rather than clustering on favorites — but weight toward the S/A tiers when the goal is cold net-new reach.\n\n**For iOS-native reveal video ads** — iMessage chat reveals (scripted thread unfolds bubble-by-bubble: screenshot hook → friend asks \"what app is that?\" → brand + promo code reveal → end card), ChatGPT reveals (typed question → streaming answer), Apple Notes reveals (a confessional note typed live), and AirDrop reveals (an incoming share where the accept-tap is the reveal) — see [references/imessage-video-ads.md](references/imessage-video-ads.md) for surface selection, the six concept angles, script and pacing rules, production routes (Playwright + ffmpeg pipeline, Remotion), craft details that sell the illusion, and the grounding/compliance rules for dramatized conversations (strictest for fabricated AI answers).\n\n**For faceless motion-style video ads** — fully generated 15–45s concept/explainer videos (styled poster stills → image-to-video \"living\" motion → TTS narration → word-timed captions; roughly $3–6 and ~15 minutes per finished video) — see [references/motion-video-ads.md](references/motion-video-ads.md) for the provider-agnostic pipeline, a nine-style visual library with fill-in prompt formulas — five characterful looks (screen-print collage, flat vector explainer, papercraft diorama, pop-art comic, claymation) plus four brand-flexible token-driven styles (monoline editorial, Swiss typographic, wireglow, duotone screenprint) driven by a brand-slots contract (FIELD / INK / ACCENT / TYPE FEEL) — the motion prompt formula, and hard-earned QC gotchas (maker-hands intrusion, final-two-seconds drift, caption/label collision, TTS/whisper sound-alikes).\n\n**For creator/UGC short-form video** — a tiered format library (reaction+demo hard cuts, \"no yapping\" split-screen tutorials, greenscreen reactions, plus Yapper, amateur investigation, David & Goliath, authority, VSL, green-screen commentary, conversation, duet/reaction, ASMR, and street-interview formats, each with a scale-vs-support tier and mechanics) and founder / organic-vlog structures (hero's journey, math, shiny-object, niche-guide, the three-capture shooting system, and the 0.5–1s cut formula) for TikTok/Reels/Shorts growth and paid — see [references/short-form-video-specs.md](references/short-form-video-specs.md). It also carries the **vertical video production spec** that applies to *all* 9:16 video this skill makes: the cross-platform safe-zone band (720×1200 text-safe area — the most-missed constraint), the classic TikTok caption recipe (white fill + black stroke, no pill), static-caption auto-sizing, and the organic-vs-baked-music decision that affects reach. Load it before producing any vertical video.\n\nFor image and video generation tools, see [references/generative-tools.md](references/generative-tools.md) for the complete guide covering:\n\n- **Image generation** — Nano Banana Pro (Gemini), Flux, Ideogram for static ad images\n- **Video generation** — Veo, Kling, Runway, Seedance, Higgsfield for video ads\n- **Voice & audio** — ElevenLabs, OpenAI TTS, Cartesia for voiceovers, cloning, multilingual\n- **Code-based video** — Remotion for templated, data-driven video at scale\n- **Platform image specs** — Correct dimensions for every ad placement\n- **Cost comparison** — Pricing for 100+ ad variations across tools\n\n**Recommended workflow for scaled production:**\n1. Generate hero creative with AI tools (exploratory, high-quality)\n2. Build Remotion templates based on winning patterns\n3. Batch produce variations with Remotion using data feeds\n4. Iterate — AI for new angles, Remotion for scale\n\n---\n\n## Generating Ad Copy\n\n### Step 1: Define Your Angles\n\nBefore writing individual headlines, establish 3-5 distinct **angles** — different reasons someone would click. Each angle should tap into a different motivation.\n\n**Common angle categories:**\n\n| Category | Example Angle |\n|----------|---------------|\n| Pain point | \"Stop wasting time on X\" |\n| Outcome | \"Achieve Y in Z days\" |\n| Social proof | \"Join 10,000+ teams who...\" |\n| Curiosity | \"The X secret top companies use\" |\n| Comparison | \"Unlike X, we do Y\" |\n| Urgency | \"Limited time: get X free\" |\n| Identity | \"Built for [specific role/type]\" |\n| Contrarian | \"Why [common practice] doesn't work\" |\n\n### Step 2: Generate Variations per Angle\n\nFor each angle, generate multiple variations. Vary:\n- **Word choice** — synonyms, active vs. passive\n- **Specificity** — numbers vs. general claims\n- **Tone** — direct vs. question vs. command\n- **Structure** — short punch vs. full benefit statement\n\n### Step 3: Validate Against Specs\n\nBefore delivering, check every piece of creative against the platform's character limits. Flag anything that's over and provide a trimmed alternative.\n\n### Step 4: Organize for Upload\n\nPresent creative in a structured format that maps to the ad platform's upload requirements.\n\n---\n\n## Iterating from Performance Data\n\nWhen the user provides performance data, follow this process:\n\n### Step 1: Analyze Winners\n\nLook at the top-performing creative (by CTR, conversion rate, or ROAS — ask which metric matters most) and identify:\n\n- **Winning themes** — What topics or pain points appear in top performers?\n- **Winning structures** — Questions? Statements? Commands? Numbers?\n- **Winning word patterns** — Specific words or phrases that recur?\n- **Character utilization** — Are top performers shorter or longer?\n\n### Step 2: Analyze Losers\n\nLook at the worst performers and identify:\n\n- **Themes that fall flat** — What angles aren't resonating?\n- **Common patterns in low performers** — Too generic? Too long? Wrong tone?\n\n### Step 3: Generate New Variations\n\nCreate new creative that:\n- **Doubles down** on winning themes with fresh phrasing\n- **Extends** winning angles into new variations\n- **Tests** 1-2 new angles not yet explored\n- **Avoids** patterns found in underperformers\n\n### Step 4: Document the Iteration\n\nTrack what was learned and what's being tested:\n\n```\n## Iteration Log\n- Round: [number]\n- Date: [date]\n- Top performers: [list with metrics]\n- Winning patterns: [summary]\n- New variations: [count] headlines, [count] descriptions\n- New angles being tested: [list]\n- Angles retired: [list]\n```\n\n---\n\n## Writing Quality Standards\n\n### Headlines That Click\n\n**Strong headlines:**\n- Specific (\"Cut reporting time 75%\") over vague (\"Save time\")\n- Benefits (\"Ship code faster\") over features (\"CI/CD pipeline\")\n- Active voice (\"Automate your reports\") over passive (\"Reports are automated\")\n- Include numbers when possible (\"3x faster,\" \"in 5 minutes,\" \"10,000+ teams\")\n\n**Avoid:**\n- Jargon the audience won't recognize\n- Claims without specificity (\"Best,\" \"Leading,\" \"Top\")\n- All caps or excessive punctuation\n- Clickbait that the landing page can't deliver on\n- Shortened or nicknamed product names — use the exact product name the\n brief gives, in every field of every concept, unless a short form is\n explicitly approved\n\n### Test Hypotheses That Can Lose\n\nWhen a concept ships with a test hypothesis, write one that a result can\nprove wrong. \"This will resonate with renters\" cannot lose; \"beats the\ncontrol on CTR among renters\" can.\n\n- Name the metric (CTR, CVR, CPA, thumb-stop rate) and the comparison —\n against which control or sibling concept, in which audience\n- State the expected direction, and a magnitude when a baseline exists\n (\"+15% CTR vs. Concept A\")\n- The hypothesis should follow from the angle's motivation, so a win or a\n loss teaches you something about the audience, not just the asset\n\n### No AI Tells\n\nAd copy that reads as generated looks like every other ad in the feed, and generating at volume multiplies whatever tic the first draft has. Check the whole batch.\n\nNever write these:\n- **Contrast reveals**: \"It's not X, it's Y.\" \"Not because X. Because Y.\" State Y directly, with the reason.\n- **Negation lists**: \"No X, no Y, no Z.\" Say what does happen. One plain absence (\"No card required\") is fine.\n- **Trailing pile-ons**: a full claim, then a comma and more restating clauses. End the sentence at the claim.\n- **Self-answered questions and colon reveals**: \"The result? 3x faster.\" \"The best part: it learns.\" FAQ questions and a question in the reader's own voice are fine.\n- **Stock phrases**: \"Say goodbye to,\" \"X, reimagined,\" \"Unlock the power of,\" \"Take it to the next level,\" \"Here's the thing.\"\n- **Em dashes** in headlines, descriptions, and primary text.\n\nAd-specific tells:\n- **Stacked negations to fit a character limit**: \"No setup. No fees. No contracts.\" Keep the one that matters most and use the space for a fact.\n- **Hook slop**: a generic \"POV:\", \"Stop scrolling,\" \"Here's what nobody tells you about X.\" Earn the next second with something specific to the viewer. A POV hook is fine when the situation is hyper-specific (\"POV: it's 3pm and you're on your fourth coffee\").\n- **CTA slop**: \"Get started today,\" \"Join thousands of happy customers.\" Name the action and what they get.\n\nAcross a batch, vary the sentence shapes. Ten variations that all open the same way read as generated, even when each one is fine alone. For the full blacklist, use the **copywriting** skill's AI-tells reference.\n\n### Descriptions That Convert\n\nDescriptions should complement headlines, not repeat them. Use descriptions to:\n- Add proof points (numbers, testimonials, awards)\n- Handle objections (\"No credit card required,\" \"Free forever for small teams\")\n- Reinforce CTAs (\"Start your free trial today\")\n- Add urgency when genuine (\"Limited to first 500 signups\")\n\n---\n\n## Output Formats\n\n### Standard Output\n\nOrganize by angle, with character counts:\n\n```\n## Angle: [Pain Point — Manual Reporting]\n\n### Headlines (30 char max)\n1. \"Stop Building Reports by Hand\" (29)\n2. \"Automate Your Weekly Reports\" (28)\n3. \"Reports Done in 5 Min, Not 5 Hr\" (31) <- OVER LIMIT, trimmed below\n -> \"Reports in 5 Min, Not 5 Hrs\" (27)\n\n### Descriptions (90 char max)\n1. \"Marketing teams save 10+ hours/week with automated reporting. Start free.\" (73)\n2. \"Connect your data sources once. Get automated reports forever. No code required.\" (80)\n```\n\n### Bulk CSV Output\n\nWhen generating at scale (10+ variations), offer CSV format for direct upload:\n\n```csv\nheadline_1,headline_2,headline_3,description_1,description_2,platform\n\"Stop Manual Reporting\",\"Automate in 5 Minutes\",\"Join 10K+ Teams\",\"Save 10+ hrs/week on reports. Start free.\",\"Connect data sources once. Reports forever.\",\"google_ads\"\n```\n\n### Static Batch Output (Mode 3)\n\nFor scaled static batches, save to a dated folder with an index:\n\n```\noutputs/YYYY-MM-DD/\n INDEX.md # every concept: template type + grounding source, scannable in 2 min\n concepts/ # one .md per concept: headline, body, visual description, image prompt, grounding\n images/ # generated images, if an image tool is configured\n```\n\nPer-concept format is defined in [references/static-ad-templates.md](references/static-ad-templates.md). The human workflow this supports: open the folder, scan INDEX.md, pick the best 5-10 for testing — picking 5 winners from 50 concepts yields better creative than picking 5 from 10.\n\n### Creative Review Page (client / stakeholder approval)\n\nWhen a person who isn't you needs to review and pick — a client, a partner, a stakeholder — produce a **creative review page**: a self-contained HTML artifact that presents each concept as an in-feed platform mockup (Instagram/Facebook, with a whitelist-handle toggle), breaks carousels into a labeled frame-by-frame storyboard, lets them toggle headline/copy variations, and discloses what's grounded in real assets. It's the visual upgrade to INDEX.md — a decision made off one link instead of by reading markdown. The template ships at [assets/creative-review-template.html](assets/creative-review-template.html) (one file, no build, hostable anywhere); populate its `DATA` object from your generated concepts. Full data model, grounding rules (the disclosure block is required), and delivery in [references/creative-review-page.md](references/creative-review-page.md).\n\n### Iteration Report\n\nWhen iterating, include a summary:\n\n```\n## Performance Summary\n- Analyzed: [X] headlines, [Y] descriptions\n- Top performer: \"[headline]\" — [metric]: [value]\n- Worst performer: \"[headline]\" — [metric]: [value]\n- Pattern: [observation]\n\n## New Creative\n[organized variations]\n\n## Recommendations\n- [What to pause, what to scale, what to test next]\n```\n\n---\n\n## Batch Generation Workflow\n\nFor large-scale creative production (Anthropic's growth team generates 100+ variations per cycle):\n\n### 1. Break into sub-tasks\n- **Headline generation** — Focused on click-through\n- **Description generation** — Focused on conversion\n- **Primary text generation** — Focused on engagement (Meta/LinkedIn)\n\n### 2. Generate in waves\n- Wave 1: Core angles (3-5 angles, 5 variations each)\n- Wave 2: Extended variations on top 2 angles\n- Wave 3: Wild card angles (contrarian, emotional, specific)\n\n### 3. Quality filter\n- Remove anything over character limit\n- Remove duplicates or near-duplicates\n- Flag anything that might violate platform policies\n- Ensure headline/description combinations make sense together\n\n---\n\n## Common Mistakes\n\n- **Writing headlines that only work together** — RSA headlines get combined randomly\n- **Ignoring character limits** — Platforms truncate without warning\n- **All variations sound the same** — Vary angles, not just word choice\n- **No CTA headlines** — RSAs need action-oriented headlines to drive clicks; include at least 2-3\n- **Generic descriptions** — \"Learn more about our solution\" wastes the slot\n- **Iterating without data** — Gut feelings are less reliable than metrics\n- **Generating without grounding** — Ungrounded concepts read like every other ad in the feed; feed the skill winning ads, reviews, and comments first\n- **Skipping the comments input** — Ad comments hold the objections and angles customers raise themselves; those usually convert best\n- **Testing too many things at once** — Change one variable per test cycle\n- **Retiring creative too early** — Allow 1,000+ impressions before judging\n\n---\n\n## Tool Integrations\n\nFor pulling performance data and managing campaigns, see the [tools registry](https://github.com/coreyhaines31/marketingskills/blob/main/tools/REGISTRY.md).\n\n| Platform | Pull Performance Data | Manage Campaigns | Guide |\n|----------|:---------------------:|:----------------:|-------|\n| **Google Ads** | `google-ads campaigns list`, `google-ads reports get` | `google-ads campaigns create` | [google-ads.md](https://github.com/coreyhaines31/marketingskills/blob/main/tools/integrations/google-ads.md) |\n| **Meta Ads** | `meta-ads insights get` | `meta-ads campaigns list` | [meta-ads.md](https://github.com/coreyhaines31/marketingskills/blob/main/tools/integrations/meta-ads.md) |\n| **LinkedIn Ads** | `linkedin-ads analytics get` | `linkedin-ads campaigns list` | [linkedin-ads.md](https://github.com/coreyhaines31/marketingskills/blob/main/tools/integrations/linkedin-ads.md) |\n| **TikTok Ads** | `tiktok-ads reports get` | `tiktok-ads campaigns list` | [tiktok-ads.md](https://github.com/coreyhaines31/marketingskills/blob/main/tools/integrations/tiktok-ads.md) |\n\n### Workflow: Pull Data, Analyze, Generate\n\n```bash\n# 1. Pull recent ad performance\nnode tools/clis/google-ads.js reports get --type ad_performance --date-range last_30_days\n\n# 2. Analyze output (identify top/bottom performers)\n# 3. Feed winning patterns into this skill\n# 4. Generate new variations\n# 5. Upload to platform\n```\n\n> **Skip the plumbing.** Step 1 is the hard part to run yourself — the Google\n> Ads (or Meta / LinkedIn / TikTok) API, OAuth, and a reporting query per\n> account, repeated every time. [Magister](https://magistermarketing.com?ref=marketingskills)\n> is an autonomous marketing agent that runs on these exact skills and pulls\n> ad performance natively: its agent reads your accounts through typed,\n> approval-gated integrations with no setup, then applies this skill to the\n> result. If you're wiring these pulls up by hand, it's worth a look.>\n> *Disclosure: Magister is built by Corey Haines, who maintains this repository.*\n> To run the pull yourself instead, use the repo's ad platform CLIs (setup in\n> [tools/SETUP.md](https://github.com/coreyhaines31/marketingskills/blob/main/tools/SETUP.md)),\n> or a reporting connector such as Supermetrics or Coupler.\n\n---\n\n## Related Skills\n\n- **ads**: For campaign strategy, targeting, budgets, and optimization\n- **marketing-loops**: For running static batch generation on a recurring cadence (the daily-creative-drop loop)\n- **customer-research**: For mining reviews and comments when building the grounded inputs corpus\n- **copywriting**: For landing page copy (where ad traffic lands)\n- **ab-testing**: For structuring creative tests with statistical rigor\n- **marketing-psychology**: For psychological principles behind high-performing creative\n- **copy-editing**: For polishing ad copy before launch\n","skills/ad-creative/evals/evals.json":"{\n \"skill_name\": \"ad-creative\",\n \"evals\": [\n {\n \"id\": 1,\n \"prompt\": \"Generate ad creative for our Meta (Facebook/Instagram) campaign. We sell an AI writing assistant for content marketers. Main value prop: write blog posts 5x faster. Target audience: content marketing managers at B2B SaaS companies. Budget: $5k/month.\",\n \"expected_output\": \"Should check for product-marketing.md first. Should generate creative following the angle-based approach: identify 3-5 angles (speed, quality, ROI, pain of blank page, competitive edge). For each angle, should generate primary text (≤125 chars), headline (≤40 chars), and description (≤30 chars) respecting Meta character limits. Should provide multiple variations per angle. Should suggest image/visual direction for each. Should organize output with angle name, hook, body, CTA for each variation. Should recommend which angles to test first.\",\n \"assertions\": [\n \"Checks for product-marketing.md\",\n \"Uses angle-based generation approach\",\n \"Identifies multiple angles (3-5)\",\n \"Respects Meta character limits (125/40/30)\",\n \"Generates multiple variations per angle\",\n \"Suggests image or visual direction\",\n \"Includes hook, body, and CTA for each\",\n \"Recommends which angles to test first\"\n ],\n \"files\": []\n },\n {\n \"id\": 2,\n \"prompt\": \"I need Google Ads copy for our CRM product. We're targeting the keyword 'best CRM for small business'. Need responsive search ads.\",\n \"expected_output\": \"Should generate Google RSA creative respecting character limits: headlines (≤30 chars each, need 10-15 variations) and descriptions (≤90 chars each, need 4+ variations). Should note that pinning should be used sparingly as it reduces optimization. Should include the target keyword in headlines. Should provide multiple angle-based variations. Should suggest ad extensions (sitelinks, callouts, structured snippets). Should follow Google Ads best practices for RSA.\",\n \"assertions\": [\n \"Respects Google RSA character limits (30 char headlines, 90 char descriptions)\",\n \"Generates 10-15 headline variations\",\n \"Generates 4+ description variations\",\n \"Includes target keyword in headlines\",\n \"Notes pinning should be used sparingly per skill guidance\",\n \"Suggests ad extensions\",\n \"Uses angle-based variation approach\"\n ],\n \"files\": []\n },\n {\n \"id\": 3,\n \"prompt\": \"Here's our ad performance data: Ad A (pain point angle) - CTR 2.1%, CPC $3.20, Conv rate 4.5%. Ad B (social proof angle) - CTR 1.4%, CPC $4.10, Conv rate 6.2%. Ad C (feature angle) - CTR 0.8%, CPC $5.50, Conv rate 2.1%. Help me iterate on these.\",\n \"expected_output\": \"Should activate the iteration-from-performance mode (not generate-from-scratch). Should analyze the data: Ad A has best CTR, Ad B has best conversion rate (highest efficiency despite lower CTR), Ad C is underperforming on all metrics. Should recommend doubling down on the pain point angle (high CTR) and social proof angle (high conversion), while pausing or reworking the feature angle. Should generate new variations that combine winning elements (pain point hook + social proof). Should suggest specific iterations on Ad A and Ad B.\",\n \"assertions\": [\n \"Activates iteration mode based on performance data\",\n \"Analyzes CTR, CPC, and conversion rate for each ad\",\n \"Identifies winning angles from the data\",\n \"Recommends pausing or reworking underperforming creative\",\n \"Generates new variations combining winning elements\",\n \"Provides specific iterations on top performers\"\n ],\n \"files\": []\n },\n {\n \"id\": 4,\n \"prompt\": \"we need linkedin ads for our enterprise security product. audience is CISOs and IT directors.\",\n \"expected_output\": \"Should trigger on casual phrasing. Should generate LinkedIn ad creative respecting character limits: introductory text (≤150 chars), headline (≤70 chars), description (≤100 chars). Should adapt tone and messaging for enterprise security audience (CISOs, IT directors) — more formal, compliance-focused, risk-reduction language. Should provide multiple angles relevant to security buyers (risk reduction, compliance, incident response time, cost of breaches). Should suggest ad format recommendations for LinkedIn (sponsored content, message ads, etc.).\",\n \"assertions\": [\n \"Triggers on casual phrasing\",\n \"Respects LinkedIn character limits (150/70/100)\",\n \"Adapts tone for enterprise security audience\",\n \"Uses risk-reduction and compliance language\",\n \"Provides multiple angles relevant to security buyers\",\n \"Suggests LinkedIn ad format recommendations\"\n ],\n \"files\": []\n },\n {\n \"id\": 5,\n \"prompt\": \"I need to generate a big batch of ad variations for a multi-platform campaign launching next week. We're a meal delivery service targeting busy professionals. Need ads for Google, Meta, and TikTok.\",\n \"expected_output\": \"Should activate the batch generation workflow. Should generate creative for all three platforms respecting each platform's character limits: Google RSA (30/90), Meta (125/40/30), TikTok (80 chars recommended, 100 max). Should identify 3-5 angles that work across platforms (convenience, health, time savings, variety, cost vs eating out). Should generate variations per angle per platform. Should note platform-specific creative considerations (TikTok needs video concepts, not just text). Should organize output clearly by platform.\",\n \"assertions\": [\n \"Activates batch generation workflow\",\n \"Generates for all three platforms\",\n \"Respects each platform's character limits\",\n \"Identifies angles that work across platforms\",\n \"Notes TikTok needs video concepts\",\n \"Organizes output by platform\",\n \"Generates multiple variations per angle per platform\"\n ],\n \"files\": []\n },\n {\n \"id\": 6,\n \"prompt\": \"Help me plan our overall paid advertising strategy. We have a $20k monthly budget and want to figure out which platforms to use and how to allocate spend.\",\n \"expected_output\": \"Should recognize this is a paid advertising strategy task, not ad creative generation. Should defer to or cross-reference the ads skill, which handles campaign strategy, platform selection, and budget allocation. May briefly mention creative considerations but should make clear that ads is the right skill for strategy.\",\n \"assertions\": [\n \"Recognizes this as paid ads strategy, not creative generation\",\n \"References or defers to ads skill\",\n \"Does not attempt full campaign strategy using creative generation patterns\"\n ],\n \"files\": []\n },\n {\n \"id\": 7,\n \"prompt\": \"I want to make one of those iMessage-style video ads for Meta — the ones where a fake text conversation reveals the product and a promo code. We sell a sleep tracking ring. Our promo code is RESTED.\",\n \"expected_output\": \"Should load references/imessage-video-ads.md. Should start by picking a concept angle from the six-angle catalog (result-as-screenshot, setup flex, cancellation moment, feature-as-punchline, friend-asks-friend inverse, receipt-as-hook) before writing bubbles — likely result-as-screenshot (a sleep score) for this product. Should draft an 8-14 bubble script in real texting voice where the brand appears only after the peer asks, with the RESTED code delivered conversationally inside a bubble and repeated on a static end card. Should apply grounding rules: any sleep-improvement claim in the thread must trace to a real customer result or product fact, and the thread must not be framed as a real testimonial. Should present production route options (a code-based Playwright+ffmpeg pipeline, or Remotion) rather than assuming one, and mention key craft rules (the recognizable send/receive SFX, silent typing indicators, 9:16 1080x1920).\",\n \"assertions\": [\n \"Loads or applies the imessage-video-ads reference\",\n \"Selects a concept angle before writing the script\",\n \"Script is 8-14 bubbles in authentic texting voice\",\n \"Brand name appears only after the peer asks about it\",\n \"Promo code RESTED appears in a bubble and on the end card\",\n \"Applies grounding rules — no fabricated claims, not framed as a real testimonial\",\n \"Mentions at least one production route and key craft rules (SFX, silent typing indicator, 9:16)\"\n ],\n \"files\": []\n },\n {\n \"id\": 8,\n \"prompt\": \"We sell a menopause supplement. I saw those ads where someone asks ChatGPT a health question and the answer recommends the product — make one of those for us. Also curious about the Apple Notes version.\",\n \"expected_output\": \"Should load references/imessage-video-ads.md and apply the Other iOS-Native Reveal Surfaces section. Should flag the compliance constraint prominently BEFORE drafting: a fabricated AI answer making health claims is the highest-risk version of this format — every claim needs substantiation, health/medical advice in a fake ChatGPT answer needs legal review, and the exchange must not be presented as a real unprompted ChatGPT output endorsing the product. May propose a compliant angle (mechanism education grounded in documented facts) or steer to the Apple Notes confession format as the lower-risk fit for a transformation story. For the Notes version: title-as-hook, first-person list with the product as the least enthusiastic line, keyboard-taps-only audio, grounding realizations in real reviews. Should apply surface-selection guidance rather than treating the three formats as interchangeable.\",\n \"assertions\": [\n \"Applies the iOS-native reveal surfaces section of the imessage-video-ads reference\",\n \"Flags health-claim/substantiation risk for the fabricated ChatGPT answer before or while drafting\",\n \"Does not present the ChatGPT exchange as a real unprompted output endorsing the product\",\n \"Recommends legal review or a compliant reframe for health advice in the AI answer\",\n \"Apple Notes guidance: title-as-hook, first-person confession, product as an understated list item, keyboard-taps-only audio\",\n \"Grounds claims and realizations in documented facts/reviews (Grounded Inputs)\",\n \"Gives surface-selection reasoning (ChatGPT vs Notes) instead of treating formats as interchangeable\"\n ],\n \"files\": []\n },\n {\n \"id\": 9,\n \"prompt\": \"Our Meta account is stuck — we've tested 30 ads over two months and nothing beats the control. I have our reviews exported and access to our ad account data. Build me a creative plan for next month.\",\n \"expected_output\": \"Should apply Mode 4 / references/creative-roadmap.md rather than jumping straight to generating ads. Should identify the account as exploration state (nothing working) and shape the plan accordingly: mostly net-new concepts across different segments/angles, minimal iterations, per-metric win redefinition (a hold-rate lift or CPC drop counts as a hit worth pulling on). Should synthesize the three signals (account performance from the ad data, customer language from the reviews, external organic — asking for or mining niche organic content) into concepts ranked by evidence tier, each with a cited source. Should produce a capacity-checked monthly slate with production tiers (favoring T1/T2 low-fidelity tests per the fidelity ladder) and flag the common exploration-state root causes to check (boring creative, overcomplicated message, unclear UVP, punishing CPMs). Should end with the retro plan for judging the slate at month end. Should not invent customer language or claims — insights must trace to the provided reviews/data.\",\n \"assertions\": [\n \"Applies the creative strategy loop (Mode 4) instead of only generating ad copy\",\n \"Diagnoses exploration state and recommends a wide, net-new-heavy mix with minimal iterations\",\n \"Redefines wins per-metric for a stuck account\",\n \"Synthesizes all three signal sources or explicitly requests the missing one\",\n \"Concepts are evidence-ranked with cited sources (no invented insights)\",\n \"Monthly slate is capacity-checked and production-tiered, favoring low-fidelity tests\",\n \"Includes a month-end retro plan that feeds the next slate\"\n ],\n \"files\": []\n },\n {\n \"id\": 10,\n \"prompt\": \"We generated four ad concepts for a client (an organic skincare brand) and need to send them something they can actually look at and approve — with the Instagram preview, the carousel frames, and the different headline options they can compare. Can you put that together?\",\n \"expected_output\": \"Should recognize this as a creative review page request and apply references/creative-review-page.md + the assets/creative-review-template.html template rather than producing plain markdown. Should copy the template into the output folder and populate its DATA object with the four concepts as tabs, each with an in-feed Instagram preview, a labeled frame-by-frame storyboard (frames labeled by narrative job — Hook / Problem / Proof / Ask — not by pictured content), selectable headline variations, primary text, and destination/CTA. Should curate to a reviewable number of concepts (2-4) rather than dumping everything. Should include a required grounding disclosure per concept stating what is real (product photography, any claims/results) and label illustrative proof as illustrative — never present invented stats or stock imagery as the brand's own. Should use styled placeholders for frames not yet rendered to image, and keep image paths relative. Should explain how to deliver it (open locally, host on a static host, or hand off the file).\",\n \"assertions\": [\n \"Produces a creative review page from the HTML template, not plain markdown\",\n \"Populates the DATA object (concept tabs, in-feed preview, frame storyboard, headline variations, copy, destination)\",\n \"Labels storyboard frames by narrative job rather than by pictured content\",\n \"Includes a required grounding/disclosure line per concept; labels illustrative proof as illustrative\",\n \"Does not present invented stats or stock imagery as the brand's real assets\",\n \"Uses placeholders for unrendered frames and keeps image paths relative\",\n \"Explains how to deliver the page (open locally / host / hand off the file)\"\n ],\n \"files\": []\n },\n {\n \"id\": 11,\n \"prompt\": \"I want to make one of those AirDrop-style video ads — where a phone gets an incoming AirDrop and you tap accept. We sell a limited-run sneaker drop.\",\n \"expected_output\": \"Should apply the AirDrop surface in references/imessage-video-ads.md (the iOS-native reveal family), not treat it as a novel format. Should build the ad around the interaction: an incoming AirDrop card (translucent sheet, sender device name, a preview thumbnail, gray Decline / blue Accept) from the receiver's POV, with the Accept tap as the reveal beat and the transfer progress-ring as the signature motion. Should make the preview thumbnail earn the tap (the sneaker money-shot / the drop), cast a relatable human sender name rather than the brand, use the AirDrop swoosh sound (not iMessage tritones) with the Apple trade-dress note, and keep it short. Should apply the family grounding/disclosure rules (a dramatization of a share, not a real endorsement; claims substantiated). May note receiver-POV-by-default vs sender-POV-as-flex.\",\n \"assertions\": [\n \"Applies the AirDrop iOS-native-reveal surface, not a from-scratch format\",\n \"Builds around the incoming-AirDrop-card + accept-tap-as-reveal interaction (receiver POV)\",\n \"Preview thumbnail is treated as the hook that must earn the accept\",\n \"Casts a relatable human sender name, not the brand, on the incoming card\",\n \"Uses the AirDrop swoosh sound + Apple trade-dress note, not iMessage tritones\",\n \"Applies the family grounding/disclosure rules (dramatized share, substantiated claims, not a real endorsement)\"\n ],\n \"files\": []\n },\n {\n \"id\": 12,\n \"prompt\": \"We're a mobile app and want to make TikTok/Reels ads. Give me a UGC reaction ad concept and make sure it won't get cut off by the app UI. Also — should we add music?\",\n \"expected_output\": \"Should load references/short-form-video-specs.md and deliver both the format and the spec. Format: the Reaction + Demo hard-cut structure (creator reaction ~3s with a hook caption written as inner monologue, hard cut to the app demo, optional payoff caption) — may also mention the other two creator formats (no-yapping split-screen, greenscreen reaction) as alternatives. Safe zone: keep all captions/key visuals inside the 720x1200 centered safe band (220px top / 500px bottom / 180px sides clear) so platform UI doesn't cover them, and use the static white-fill/black-stroke caption style that auto-sizes to fit. Music: give the organic-vs-baked decision — for organic posting, export without baked music and attach the trending sound in-app (algorithm reward); bake music only for paid ads or where native sound can't be attached, fading out the last ~0.8s.\",\n \"assertions\": [\n \"Provides the reaction+demo hard-cut structure with the hook caption as the reaction's inner monologue\",\n \"Specifies the cross-platform safe band (roughly 220 top / 500 bottom / 180 sides, or the 720x1200 text-safe area) so captions aren't covered by platform UI\",\n \"Describes the static white-fill/black-stroke caption style with auto-sizing (no animated captions)\",\n \"Gives the organic-vs-baked-music decision rather than a blanket yes/no (attach trending sound in-app for organic; bake for ads)\"\n ],\n \"files\": []\n },\n {\n \"id\": 13,\n \"prompt\": \"We're a DTC brand with a stalled Meta account and need fresh static ad concepts that can actually open cold net-new audiences — not just retarget. Which static templates should we lead with, and which should we avoid right now? Also, we have several SKUs.\",\n \"expected_output\": \"Should load references/static-ad-templates.md and reason from the tier + funnel-role tagging rather than treating all templates as interchangeable. For cold net-new reach, should prioritize the S/A-tier statics — Founder Message and Origin Story (S, founder content is the reliable first cold-scaler) and, because the brand has multiple SKUs, the Grid Static (A, multi-SKU/bundle, low-hanging fruit that scales cold). Should explain the unicorn-scaler-vs-supporting-cast lens: most B-tier templates (Us vs. Them, Before/After, FAQ Card, Callout) convert mid-funnel and shouldn't be expected to open cold reach or be killed for failing to. Should flag the decayed formats to avoid: Press Mention (F — rights nightmare), Testimonial statics (E — unless golden-nugget), Numbered List/Listicle (E — dead lately). Should keep grounding rules (concepts trace to real reviews/winning ads/comments; no fabricated social proof). May cross-reference the fuller format map for video/partnership formats.\",\n \"assertions\": [\n \"Loads or applies the static-ad-templates reference and reasons from tier + funnel role\",\n \"Prioritizes S/A-tier statics for cold reach (Founder Message, Origin Story, Grid Static)\",\n \"Recommends the Grid Static specifically given multiple SKUs\",\n \"Explains the unicorn-scaler vs. supporting-cast lens (B-tier = mid-funnel, don't kill for failing to scale cold)\",\n \"Flags decayed formats to avoid (Press Mention F, Testimonial statics E, Listicle/Numbered List E)\",\n \"Preserves grounding rules — no fabricated social proof\"\n ],\n \"files\": []\n },\n {\n \"id\": 14,\n \"prompt\": \"We're a DTC supplement brand and our Meta reach has been flat for weeks. We can make basically any ad. What creative format should we make next, and what should we NOT waste time on?\",\n \"expected_output\": \"Should load references/meta-creative-formats.md and answer as a which-format-to-make-next decision, not a from-scratch copy dump. Should lead with the unicorn-scaler vs. supporting-cast lens and the persona-based Andromeda context (creator-fronted formats reach personas natively), and tie the flat/declining reach specifically to deploying creator-fronted formats — especially partnership ads (the #1 priority) — to restore net-new reach. Should surface the S-tier picks (founder content as the reliable first winner, partnership ads, VSL for education-heavy niches like supplements) and relevant A-tier options (authority ads fit a supplement brand, grid statics as low-hanging fruit). Should explicitly de-prioritize F-tier (press ads, podcast ads unless a founder is on a known show, notes-app/UX fake-native ads that 'do not convert' and confuse the algorithm). Should frame the answer as building a portfolio (scalers + supporting cast), and route to the static/video references for how to actually build the chosen format.\",\n \"assertions\": [\n \"Loads or applies the meta-creative-formats reference\",\n \"Frames the answer with the unicorn-scaler vs. supporting-cast distinction\",\n \"Explains the persona-based Andromeda reason creator-fronted formats rank highest\",\n \"Ties flat/declining reach to deploying partnership ads (the #1 priority) to restore net-new reach\",\n \"Recommends S-tier picks (founder content, partnership ads, VSL) and a fitting A-tier option (authority ads and/or grid statics)\",\n \"Explicitly de-prioritizes F-tier (press, podcast-unless-known-show, notes-app/UX fake-native)\",\n \"Frames it as building a portfolio and routes to static/video references for production\"\n ],\n \"files\": []\n },\n {\n \"id\": 15,\n \"prompt\": \"We're a health supplement brand and want video ads that will actually scale to cold audiences, not just retarget. What creator formats should we prioritize, and can our founder be in them?\",\n \"expected_output\": \"Should load references/short-form-video-specs.md and reason from the scale-vs-support tier logic, not list formats flatly. For scaling cold in a trust-gated health niche it should prioritize the higher-tier creator-fronted formats — VSL (S; upfront education, the mechanism-then-offer script) and Authority (A; a credentialed expert, with the caveat that health claims must be real/substantiated and routed through legal review per Grounded Inputs) — and can also point to Yapper, Amateur Investigation, and David & Goliath (all A) as cold-scaling options. Founder: yes — founder's content is often a brand's first top performer, and the founder can carry a Yapper or David & Goliath via the founder/organic-vlog structures (hero's journey, math, shiny-object, niche-guide). Should mention the practical production system (three-capture close/medium/wide shooting, 0.5–1s cut formula) and frame the answer as building a portfolio across tiers rather than betting on one format.\",\n \"assertions\": [\n \"Reasons from the scale-vs-support tier logic (prioritizes higher-tier cold-scaling formats over a flat list)\",\n \"Recommends VSL and/or Authority for the education-heavy, trust-gated health niche, and flags the health-claims/legal-review compliance caveat for the Authority/expert format\",\n \"Confirms the founder can front the ads (founder content as a common first top performer) via a founder/organic-vlog structure such as hero's journey or David & Goliath\",\n \"References the founder shooting/edit system (three-capture close/medium/wide and/or the 0.5–1s cut formula) and/or framing the mix as a portfolio across tiers\"\n ],\n \"files\": []\n },\n {\n \"prompt\": \"Give me 10 Meta ad primary-text variations for our no-code website builder. It needs no developer, no hosting setup, and no credit card to start.\",\n \"expected_output\": \"Should write 10 variations that avoid AI tells even though the brief is a list of absences. Should not stack negations ('No developer. No hosting. No card.') in any variation; may state one absence plainly. No contrast reveals, self-answered questions, colon reveals, em dashes, or stock phrases ('Say goodbye to', 'Unlock'). No two variations should share the same sentence shape, and hooks should be specific rather than 'POV:' or 'Stop scrolling'.\",\n \"assertions\": [\n \"No variation stacks three negations\",\n \"No contrast reveals or self-answered questions\",\n \"No em dashes\",\n \"Variations don't repeat the same sentence shape\",\n \"No 'POV:' / 'Stop scrolling' / stock-phrase hooks\"\n ],\n \"files\": [],\n \"id\": 16\n },\n {\n \"id\": 17,\n \"prompt\": \"We sell forecasting software. Problem-aware finance teams spend Friday comparing CSV exports. Product facts: imports CSVs into one approval queue; no quantified time study. Customer notes: R1 says “three versions of the final spreadsheet”; R2 says “I want Friday to end with a finished forecast.” Existing ad A uses R1 and gets clicks but few qualified demos. Create three text-led LinkedIn static concepts and a mobile visual brief.\",\n \"expected_output\": \"Applies awareness-headline reference; distinct grounded treatments with R1/R2/product evidence, no fabricated hours/statistics. Specifies headline-dominant composition and actual mobile readability including qualifiers, not invented LinkedIn policy. Diagnoses qualified-demo continuity rather than calling CTR a win.\",\n \"assertions\": [\n \"Uses a stated problem-aware audience assumption\",\n \"Produces distinct grounded concepts with source IDs\",\n \"Does not invent quantified time-saving or fake customer dialogue\",\n \"Includes actual-size mobile headline/qualification readability check\",\n \"Connects success to qualified demos rather than clicks alone\"\n ],\n \"files\": []\n },\n {\n \"id\": 18,\n \"prompt\": \"Our meal delivery service has two customer comments: C1 “I am still on the train when my kids ask about dinner,” C2 “Sunday grocery planning starts over every week.” Product fact F1: one order supplies three dinners. Make cold-audience static ideas using conversation and a stat callout. Can you make up three more quotes so we have five?\",\n \"expected_output\": \"Uses supplied language and fact with unfamiliar/problem-aware assumptions; declines invented customer quotes and reports limited corpus. A invented exchange if used is explicitly dramatization without fake attribution; three-dinner quantity traceable to F1. Provides truthful legible layouts.\",\n \"assertions\": [\n \"Reports two available customer phrases rather than inventing five\",\n \"Labels invented dialogue as dramatization or uses permitted actual wording\",\n \"Traces quantity to F1\",\n \"Distinguishes awareness hypotheses and supplies mobile visual direction\"\n ],\n \"files\": []\n },\n {\n \"id\": 19,\n \"prompt\": \"We have a pilot summary: 8 SaaS teams, March, weekly reconciliation median fell from 6 hours to 2; no independent control. Make a headline creative saying every customer saves 67%, with a tiny footnote. We also want to shame buyers still using spreadsheets.\",\n \"expected_output\": \"Keeps result scoped to pilot median/sample/period, declines universal causal guarantee and buyer shaming. A headline with meaningful legible adjacent qualifiers or a safer non-stat alternative; qualified copy cannot be hidden in tiny text. Preserves truthful precision rather than changing number to seem real.\",\n \"assertions\": [\n \"Does not generalize pilot to every customer or causal guarantee\",\n \"Keeps sample/period/median scope available in readable qualification\",\n \"Rejects shrinking material qualification into an unreadable footnote\",\n \"Critiques workflow without shaming buyers\",\n \"Does not claim round numbers are fake\"\n ],\n \"files\": []\n },\n {\n \"id\": 20,\n \"prompt\": \"Give me six headline concepts for our product. No product description, audience, customer language, proof, platform, or offer is available. Can you just fill every formula with made-up results?\",\n \"expected_output\": \"Requests the key missing context and evidence instead of manufactured claims; may explain treatments as hypothetical structure, but not ready-to-run ungrounded product ads. Does not force six outputs from an empty corpus.\",\n \"assertions\": [\n \"Asks for missing product/audience/evidence/placement context\",\n \"Does not fabricate results or testimonials\",\n \"Keeps illustrative formulas distinct from deliverable ads\",\n \"Does not promise formula choice proves conversion\"\n ],\n \"files\": []\n }\n ]\n}\n","skills/ad-creative/references/awareness-headlines.md":"# Awareness-Based Headline Creative\n\nUse for text-led static concepts, especially B2B feed ads. These are compact treatments within the existing angle and template workflow; they do not replace audience research, the hook system, or platform specs.\n\nConcept lineage: the maintainer's #404 follow-up to the ivangfalco/ads-skills remix. Re-expressed independently and reconciled with this repo's grounding and AI-tells rules; no source copy or creative assets imported.\n\n## Choose the Reader's Next Step\n\n| Awareness | What the reader needs | Useful treatments | CTA direction |\n|---|---|---|---|\n| Unaware / unfamiliar with the category | Recognize a specific lived situation | Feeling, conversation | Learn about the situation or mechanism |\n| Problem-aware | See a credible path out of a named friction | Pain + outcome, time-anchored friction | See the workflow or explanation |\n| Solution-aware | Understand a concrete way this approach differs | State change, documented statistic | Compare the mechanism or watch a demo |\n| Product-aware | Resolve an objection with specific proof | Statistic, pain + outcome | Inspect proof, try, or book the relevant next step |\n\nStage is a working hypothesis about this audience, not a diagnosis based on the platform alone. If unknown, state the assumption and propose two distinct stage hypotheses. A formula alone cannot prove an ad will convert.\n\n## Six Treatments\n\nThe examples below are invented illustrations of structure, **not** approved product claims, real customer quotes, or ready-to-run ads. Replace their facts with the user's verified corpus. Each treatment can serve multiple stages; choose based on what the reader already knows.\n\n| Treatment | Construction and guardrail | Illustrative SaaS example | Illustrative DTC example |\n|---|---|---|---|\n| **Feeling** | Name a recognizable feeling tied to a specific work/life moment. Avoid vague empowerment or a forced positive/negative reveal. | “Leave Friday with the forecast finished.” | “Dinner sorted before the commute home.” |\n| **Conversation** | Two short lines from a recognizable situation. Use an actual permitted exchange, or label invented dialogue as a dramatization; never fabricate testimonial attribution. | “Is this the final sheet?” / “There are three final sheets.” | “What's for dinner?” / “I'm still on the train.” |\n| **State change** | Show the current and proposed workflow with concrete artifacts. State what changes; don't force “not X, Y” copy or unsupported competitor claims. | “Bring the three approval threads into one queue.” | “Pack tomorrow's lunch while tonight's dinner cooks.” |\n| **Time-anchored friction** | Name an obsolete or repetitive process and its actual moment. Critique the process without shaming the buyer or inventing a trend. | “Month-end still starts with copying CSVs.” | “The Sunday grocery list starts over again.” |\n| **Documented statistic** | One supported number plus the population, period, or meaningful qualifier. Keep its real precision; rounded numbers are not evidence of fakery. | “Pilot teams cut weekly reconciliation from 6 hours to 2.” | “Three dinners packed in one delivery.” |\n| **Pain + outcome** | One named friction and a concrete next action/outcome, usually under ten words. Brevity is a draft constraint, not permission to drop a material qualifier. | “Find the missing invoice before close.” | “Get dinner ready between calls.” |\n\nRead each draft against the existing AI-tells guidance. Do not produce six near-identical sentences just to fill a formula menu. Preserve customer language where it is permitted and relevant; facts can ground a paraphrase without becoming a fabricated quote.\n\n## Gather Proof Before Copy\n\nBuild a small evidence sheet: source ID, verbatim phrase or product fact, audience, date, allowed usage, and limitations. Aim for five useful customer phrases when available, across more than one perspective. If the available corpus is smaller, report the gap and reduce diversity claims; do not invent quotes to meet a count. Existing Grounded Inputs requirements still apply to scaled Mode 3 batches.\n\nSeparate **language evidence** (“the export takes forever”) from **claim evidence** (a measured reduction across a defined pilot). A complaint supports the pain angle; it does not substantiate a numerical performance promise. Preserve a pilot's sample size and period in adjacent readable copy or withhold the statistical concept when a truthful presentation cannot fit.\n\n## Make the Headline the Visual\n\nApply this treatment to Headline Statement or Stat Callout from the static library when the words carry the concept. It is a production hypothesis, not a guarantee that text-only ads beat every other format.\n\n- Start with one dominant sentence and a bold, legible sans-serif face. As a design starting point, let the headline occupy roughly 50–70% of the composition; adjust after actual-size testing.\n- Use intentional line breaks that keep phrases together. Keep the product/logo secondary unless recognizing the product is the concept's job.\n- Test a dark background with light text, or a strong brand-colored background with contrasting text. Choose from the user's actual palette and verify readability; avoid a color prescription based on an unverified feed theme.\n- Include the audience's specific moment, one defensible idea, and enough qualification to read truthfully. Exclude stock imagery competing with the sentence, tiny disclaimers, fake review UI, extra badges, and an unsupported second claim.\n- Make a preview at the **actual mobile feed display size**, not only on a desktop canvas. Check that a reviewer can read the entire headline and its material qualifiers without zooming. If not, shorten the claim or change the layout; never solve it by shrinking required qualifications.\n- Get current export dimensions, cropping, safe areas, and text-field limits from the platform-spec reference and the selected placement. The 50–70% layout suggestion is not a platform rule or a text-coverage approval threshold.\n\n## Deliver and Learn\n\nFor each concept, provide: audience/stage assumption, distinct angle, treatment + static template, headline, supporting qualification, CTA, source IDs, and visual specification with mobile-readability check. Mark evidence gaps and keep output as concepts for review unless publication is already authorized.\n\nTest stage/angle hypotheses before fine font changes. Compare the relevant downstream qualified outcome with the existing control, keeping offer and audience stable where possible. Strong CTR with weak qualified conversion is a reason to examine the promise and landing-page continuity, not proof of a winning headline.\n","skills/ad-creative/references/creative-review-page.md":"# The Creative Review Page\n\nA shareable, self-contained web page that presents generated ad concepts for a client or stakeholder to **review and pick** — the visual upgrade to `INDEX.md`. Where the markdown outputs are built for the operator, the review page is built for the person approving the spend: it shows each concept as an in-feed platform mockup, breaks carousels into a labeled frame-by-frame storyboard, lets them toggle copy variations, and discloses what's grounded in real assets.\n\nThe template ships at [assets/creative-review-template.html](../assets/creative-review-template.html). It's one file — inline CSS and JS, no build, no dependencies, no network. Open it locally, host it on any static host (Vercel/Netlify/GitHub Pages), or hand off the `.html` file directly.\n\n## When to produce one\n\n- **Presenting a batch for approval** — after Mode 1 or Mode 3 generation, package the top concepts into a review page instead of (or alongside) `INDEX.md`. Picking 5 of 50 is a *visual* decision; a client shouldn't have to read markdown to make it.\n- **Pitching a whitelist / co-branded partnership** — the format the source pattern was built for: show the partner exactly what the ad looks like under each handle, with the rollout mechanics spelled out.\n- **A monthly slate review** (Mode 4) — render the slate's concepts so the account-state call and the pick happen off one link.\n\nDon't produce one for a single headline tweak or a quick internal gut-check — the markdown output is faster. Reach for the review page when a human who isn't you needs to choose.\n\n## How it's built\n\nThe template renders entirely from a JSON block near the top of the file — `` would otherwise close the data block early. Since agents write the JSON, apply this escape mechanically to all string values. All values are HTML-escaped again at render time, so this is defense-in-depth, but the source-level escape is the one that matters — do it.\n\n## Producing and delivering it\n\n1. Copy `assets/creative-review-template.html` into the batch's output folder as `review.html` (e.g. `outputs/YYYY-MM-DD/review.html`).\n2. Replace the `DATA` object with the real project — concepts, frames, copy, grounding. Populate `image` paths for any frames you've rendered (keep them relative to the html file so the folder stays portable).\n3. Verify it renders: open it in a browser, click through every concept tab, both platform and handle toggles, and each frame in the storyboard.\n4. Deliver: hand off the folder (html + `images/`), or host it. For a client link, `vercel deploy` or any static host works — it's a single page with local assets.\n\nKeep the review page next to the markdown outputs, not instead of them: `INDEX.md` and the per-concept files remain the operator's record and the grounding audit trail; `review.html` is the approval surface built on top.\n\n## Common mistakes\n\n- **Too many concepts** — 2–4 tabs is a decision; 10 is a menu nobody finishes. Curate before you present.\n- **Unlabeled or content-labeled frames** — label by narrative job (`The proof`), not by what's pictured (`Table screenshot`).\n- **Missing or dishonest grounding** — every concept discloses what's real; illustrative proof is labeled illustrative.\n- **Editing the render code** — everything is data-driven; if something won't show, it's a `DATA` field, not the JS.\n- **Absolute image paths** — keep image paths relative so the output folder can be zipped, moved, or hosted intact.\n","skills/ad-creative/references/creative-roadmap.md":"# The Creative Strategy Loop\n\nGeneration (Modes 1–3) answers \"make me ads.\" This reference answers the question that comes first: **which ads are worth making, in what order, at what production cost** — and the retro that turns each month's results into next month's plan. It's the standing operating loop of a creative strategist, run by an agent with a human deciding.\n\n```\nSignals → Concepts (evidence-ranked) → Roadmap (tiered, capacity-checked) → Briefs → [Modes 1–3 produce] → Monthly retro → back into the icebox\n```\n\n---\n\n## Step 1: Read the Three Signals\n\nCreative direction comes from synthesis across three independent signal sources. One source alone misleads: the account tells you what worked *among things you've tried*, customers tell you why they buy *in their words*, and organic content tells you what the audience *chooses to watch when nobody's paying*.\n\n| Signal | What to pull | How |\n|---|---|---|\n| **Account performance** | Winners/losers by angle, hook, format; funnel metrics per concept (see [hook-system.md](hook-system.md) diagnostic funnel); fatigue state | `google-ads` / `meta-ads` / `linkedin-ads` / `tiktok-ads` CLIs (see Tool Integrations in SKILL.md) |\n| **Customer/brand** | Verbatim pain/desire/objection language; unexpected use cases; who's *actually* buying vs. who's targeted | The Grounded Inputs corpus (`inputs/reviews/`, `inputs/comments/`), sales-call notes, support themes — per **customer-research** |\n| **External organic** | What the niche watches unpaid: top organic content, its hooks, formats, vocabulary; competitor ads running long enough to be presumed working | **scraping**, the social listening tooling in **social**, ad libraries, **competitor-profiling** |\n\n**Cadence:** a monthly deep dive (60–90 min, all three sources, feeds the monthly roadmap) plus a weekly ~20-minute refresh (what changed: new winners/losers, new review themes, anything spiking organically). Research beyond what the next decision needs is busywork — every synthesis session should end in concepts, not notes.\n\n**Trust rule:** every insight the agent surfaces must carry its receipt — which review, which ad's metrics, which organic post. An insight without a source doesn't enter the icebox. (Same grounding rules as everything else in this skill.)\n\n---\n\n## Step 2: Turn Signals into Evidence-Ranked Concepts\n\nA **concept** is one testable creative hypothesis: *segment × motivation × angle × format*, with its evidence attached. \"UGC for moms\" is not a concept; \"new-parent insomniacs (per 40+ reviews mentioning 3am feeds) × 'quiet enough to not wake the baby' × before/after demo × POV night-shot video\" is.\n\nRank every concept by the strongest evidence supporting it:\n\n| Tier | Evidence | Weight |\n|---|---|---|\n| 1 | Your own account: a converting ad with the same angle/segment | Strongest — iterate and extend |\n| 2 | Your customers verbatim: recurring review/call language | Strong — build new creative on it |\n| 3 | Competitor creative running 60+ days (presumed working) | Good — adapt the angle, never the ad |\n| 4 | Organic engagement in the niche (unpaid views/saves on the theme) | Moderate — validate cheaply first |\n| 5 | Cross-niche pattern (worked in an adjacent category) | Weak — icebox until corroborated |\n| 6 | Team hunch, no external signal | Weakest — low-fi test or drop |\n\nHigher evidence earns roadmap *priority* — an earlier slot in the slate. Production tier is a separate call, set by validation strength, existing assets, capacity, and risk: even a tier-2 customer-language concept starts low-fidelity until it shows a funnel signal. Hunches aren't banned — they're just cheap and last.\n\n---\n\n## Step 3: Branch on Account State\n\nThe right creative mix depends on which of two states the account is in. Diagnose before roadmapping — a plan built for the wrong state wastes the month.\n\n**Exploration state** — nothing (or nothing new) is working:\n- Go **wide, not deep**: mostly net-new concepts across different segments and angles; keep iterations to a small minority — iterating on losers multiplies losers\n- **Redefine \"win\" per-metric**: with no full-funnel winners, a single-metric improvement (a hold-rate lift, a CPC drop, a CVR bump) on any test is a hit worth pulling on — see the diagnostic funnel\n- Iterate **only on hits**; everything else stays exploratory\n- Common root causes to check while testing: the creative is boring (safe, seen-before), the message is overcomplicated, the offer/UVP is unclear, or CPMs are punishing a too-narrow audience\n\n**Scaling state** — one or more concepts are converting profitably:\n- Go **deep on the winner** while it's open: a winner-led slate of visually-distinct variations of the winning concept (same message, new execution — near-duplicates mostly cannibalize the original's reach and teach you nothing new, so variations must look meaningfully different), plus a remix lane (tonal/emotional re-executions of it) and sub-angle probes drilling *into* the winning segment; tune the split to budget, fatigue speed, and production velocity\n- Keep a small exploration allocation alive even mid-scale — winners fatigue, and the next winner is rarely an iteration of the current one\n- Speed matters more in this state: a scaling window is finite\n\n---\n\n## Step 4: The Roadmap Artifact\n\nMaintain one living document (suggested: `roadmap.md` beside the Grounded Inputs corpus) with three horizons:\n\n```\n## Icebox — every concept, evidence tier + source attached, nothing scheduled\n## This quarter — 2-4 themes chosen from the icebox (the bets), with why-now\n## This month — the slate: concept | evidence tier | production tier | owner | status\n```\n\nEach monthly-slate concept gets a **production tier**:\n\n| Tier | Cost | What it is | Use for |\n|---|---|---|---|\n| **T1 — Iteration** | Hours | New hook/caption/crop on an existing asset | Extending proven winners |\n| **T2 — Remix** | Days | New creative from existing footage/assets/AI generation | Concepts with decent evidence or a first low-fi signal |\n| **T3 — Production** | Weeks | Net-new shoot, creators, full build | Only angles with own-account proof or a prior low-fi funnel signal (fidelity ladder in [hook-system.md](hook-system.md)) |\n\n**Capacity check — the rule that keeps roadmaps honest:** count what the team (or the AI pipeline) can produce *at quality* this month, and roadmap to that number. A 20-concept slate against 8 concepts of real capacity doesn't produce 20 ads; it produces 20 compromised ones and a burned-out team. Cut by evidence rank until the slate fits.\n\nFrom the slate, generate **one brief per concept** (segment, motivation + verbatim source, angle, format, hook matrix rows, production tier, success metric) and hand each to Modes 1–3 for production.\n\n---\n\n## Step 5: The Monthly Creative Retro\n\nLast step of the loop, first input of the next one. One artifact per month (suggested: `retros/YYYY-MM.md`):\n\n```\n## Winners — concept, the funnel numbers, and the WHY (which element earned it)\n## Losers — concept, where in the funnel it died, hypothesis for why\n## Metric wins — full-funnel losers with one strong metric (these are leads, not losses)\n## Learnings — pattern-level notes → written back into the icebox as new/revised concepts\n## Kills — concepts retired from the icebox, with reason\n## Next slate — first draft of next month, updated evidence ranks\n```\n\nRetro rules:\n\n- **Judge concepts, not ads.** Three executions of one concept failing says the concept is wrong; one failing says the execution was.\n- **Read the funnel, not the ROAS column.** The diagnostic funnel says *what* to fix; ROAS alone says only *that* something is broken.\n- **Enough data before verdicts** — respect the impression/spend thresholds in Common Mistakes and the **ads** skill's decision systems; a two-day read is a coin flip.\n- **Every learning lands somewhere**: icebox update, evidence re-rank, or kill. A retro that changes nothing in the roadmap was a meeting, not a retro.\n\nTo run this loop on a schedule (retro on the 1st, weekly refresh Mondays, daily batches via Mode 3), see the creative loops in **marketing-loops**.\n\n---\n\n## Failure Modes\n\n- **Roadmapping without a diagnosis** — a slate built before reading the three signals is a wish list; testing without a diagnosis isn't strategy\n- **Iteration-heavy slates in exploration state** — polishing losers while the real problem (angle, offer, audience) goes untested\n- **Ignoring capacity** — the plan the team can't produce at quality is a plan to produce slop\n- **Evidence-free concepts jumping the queue** — the loudest stakeholder's hunch ships as a T3 shoot while tier-2 customer language sits in the icebox\n- **Retro as theater** — winners celebrated, nothing re-ranked, icebox untouched\n- **Scaling-state complacency** — 100% of the slate on winner variations; when the winner fatigues, the pipeline is empty\n","skills/ad-creative/references/generative-tools.md":"# Generative AI Tools for Ad Creative\n\nReference for using AI image generators, video generators, and code-based video tools to produce ad visuals at scale.\n\n---\n\n## When to Use Generative Tools\n\n| Need | Tool Category | Best Fit |\n|------|---------------|----------|\n| Static ad images (banners, social) | Image generation | ChatGPT Images 2.0, Nano Banana Pro, Flux, Ideogram |\n| Ad images with text overlays | Image generation (text-capable) | Ideogram, Nano Banana Pro |\n| Short video ads (6-30 sec) | Video generation | Veo, Kling, Runway, Seedance |\n| Video ads with voiceover | Video gen + voice | Veo (native), or Runway + ElevenLabs |\n| Voiceover tracks for ads | Voice generation | ElevenLabs, OpenAI TTS, Cartesia |\n| Multi-language ad versions | Voice generation | ElevenLabs, PlayHT |\n| Brand voice cloning | Voice generation | ElevenLabs, Resemble AI |\n| Product mockups and variations | Image generation + references | Flux (multi-image reference) |\n| Templated video ads at scale | Code-based video | Remotion |\n| Personalized video (name, data) | Code-based video | Remotion |\n| Brand-consistent variations | Image gen + style refs | Flux, Ideogram, Nano Banana Pro |\n\n---\n\n## Image Generation\n\n### Nano Banana Pro (Gemini)\n\nGoogle DeepMind's image generation model, available through the Gemini API.\n\n**Best for:** High-quality ad images, product visuals, text rendering\n**API:** Gemini API (Google AI Studio, Vertex AI)\n**Pricing:** ~$0.04/image (Gemini 2.5 Flash Image), ~$0.24/4K image (Nano Banana Pro)\n\n**Strengths:**\n- Strong text rendering in images (logos, headlines)\n- Native image editing (modify existing images with prompts)\n- Available through the same Gemini API used for text generation\n- Supports both generation and editing in one model\n\n**Ad creative use cases:**\n- Generate social media ad images from text descriptions\n- Create product mockup variations\n- Edit existing ad images (swap backgrounds, change colors)\n- Generate images with headline text baked in\n\n**API example:**\n```bash\n# Using the Gemini API for image generation\ncurl -X POST \"https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash-image:generateContent\" \\\n -H \"Content-Type: application/json\" \\\n -H \"x-goog-api-key: $GEMINI_API_KEY\" \\\n -d '{\n \"contents\": [{\"parts\": [{\"text\": \"Create a clean, modern social media ad image for a project management tool. Show a laptop with a kanban board interface. Bright, professional, 16:9 ratio.\"}]}],\n \"generationConfig\": {\"responseModalities\": [\"TEXT\", \"IMAGE\"]}\n }'\n```\n\n**Docs:** [Gemini Image Generation](https://ai.google.dev/gemini-api/docs/image-generation)\n\n---\n\n### Flux (Black Forest Labs)\n\nOpen-weight image generation models with API access through Replicate and BFL's native API.\n\n**Best for:** Photorealistic images, brand-consistent variations, multi-reference generation\n**API:** Replicate, BFL API, fal.ai\n**Pricing:** ~$0.01-0.06/image depending on model and resolution\n\n**Model variants:**\n| Model | Speed | Quality | Cost | Best For |\n|-------|-------|---------|------|----------|\n| Flux 2 Pro | ~6 sec | Highest | $0.015/MP | Final production assets |\n| Flux 2 Flex | ~22 sec | High + editing | $0.06/MP | Iterative editing |\n| Flux 2 Dev | ~2.5 sec | Good | $0.012/MP | Rapid prototyping |\n| Flux 2 Klein | Fastest | Good | Lowest | High-volume batch generation |\n\n**Strengths:**\n- Multi-image reference (up to 8 images) for consistent identity across ads\n- Product consistency — same product in different contexts\n- Style transfer from reference images\n- Open-weight Dev model for self-hosting\n\n**Ad creative use cases:**\n- Generate 50+ ad variations with consistent product/person identity\n- Create product-in-context images (your SaaS on different devices)\n- Style-match to existing brand assets using reference images\n- Rapid A/B test image variations\n\n**Docs:** [Replicate Flux](https://replicate.com/black-forest-labs/flux-2-pro), [BFL API](https://docs.bfl.ml/)\n\n---\n\n### Ideogram\n\nSpecialized in typography and text rendering within images.\n\n**Best for:** Ad banners with text, branded graphics, social ad images with headlines\n**API:** Ideogram API, Runware\n**Pricing:** ~$0.06/image (API), ~$0.009/image (subscription)\n\n**Strengths:**\n- Best-in-class text rendering (~90% accuracy vs ~30% for most tools)\n- Style reference system (upload up to 3 reference images)\n- 4.3 billion style presets for consistent brand aesthetics\n- Strong at logos and branded typography\n\n**Ad creative use cases:**\n- Generate ad banners with headline text directly in the image\n- Create social media graphics with branded text overlays\n- Produce multiple design variations with consistent typography\n- Generate promotional materials without needing a designer for each iteration\n\n**Docs:** [Ideogram API](https://developer.ideogram.ai/), [Ideogram](https://ideogram.ai/)\n\n---\n\n### Other Image Tools\n\n| Tool | Best For | API Status | Notes |\n|------|----------|------------|-------|\n| **DALL-E 3** (OpenAI) | General image generation | Official API | Integrated with ChatGPT, good text rendering |\n| **Midjourney** | Artistic, high-aesthetic images | No official public API | Discord-based; unofficial APIs exist but risk bans |\n| **Stable Diffusion** | Self-hosted, customizable | Open source | Best for teams with GPU infrastructure |\n\n---\n\n## Video Generation\n\n### Google Veo\n\nGoogle DeepMind's video generation model, available through the Gemini API and Vertex AI.\n\n**Best for:** High-quality video ads with native audio, vertical video for social\n**API:** Gemini API, Vertex AI\n**Pricing:** ~$0.15/sec (Veo 3.1 Fast), ~$0.40/sec (Veo 3.1 Standard)\n\n**Capabilities:**\n- Up to 60 seconds at 1080p\n- Native audio generation (dialogue, sound effects, ambient)\n- Vertical 9:16 output for Stories/Reels/Shorts\n- Upscale to 4K\n- Text-to-video and image-to-video\n\n**Ad creative use cases:**\n- Generate short video ads (15-30 sec) from text descriptions\n- Create vertical video ads for TikTok, Reels, Shorts\n- Produce product demos with voiceover\n- Generate multiple video variations from the same prompt with different styles\n\n**Docs:** [Veo on Vertex AI](https://cloud.google.com/vertex-ai/generative-ai/docs/video/overview)\n\n---\n\n### Kling (Kuaishou)\n\nVideo generation with simultaneous audio-visual generation and camera controls.\n\n**Best for:** Cinematic video ads, longer-form content, audio-synced video\n**API:** Kling API, PiAPI, fal.ai\n**Pricing:** ~$0.09/sec (via fal.ai third-party)\n\n**Capabilities:**\n- Up to 3 minutes at 1080p/30-48fps\n- Simultaneous audio-visual generation (Kling 2.6)\n- Text-to-video and image-to-video\n- Motion and camera controls\n\n**Ad creative use cases:**\n- Longer product explainer videos\n- Cinematic brand videos with synchronized audio\n- Animate product images into video ads\n\n**Docs:** [Kling AI Developer](https://klingai.com/global/dev/model/video)\n\n---\n\n### Runway\n\nVideo generation and editing platform with strong controllability.\n\n**Best for:** Controlled video generation, style-consistent content, editing existing footage\n**API:** Runway Developer Portal\n\n**Capabilities:**\n- Gen-4: Character/scene consistency across shots\n- Motion brush and camera controls\n- Image-to-video with reference images\n- Video-to-video style transfer\n\n**Ad creative use cases:**\n- Generate video ads with consistent characters/products across scenes\n- Style-transfer existing footage to match brand aesthetics\n- Extend or remix existing video content\n\n**Docs:** [Runway API](https://docs.dev.runwayml.com/)\n\n---\n\n### Seedance 2.0 (ByteDance)\n\nByteDance's video generation model with simultaneous audio-visual generation and multimodal inputs.\n\n**Best for:** Fast, affordable video ads with native audio, multimodal reference inputs\n**API:** BytePlus (official), Replicate, WaveSpeedAI, fal.ai (third-party); OpenAI-compatible API format\n**Pricing:** ~$0.10-0.80/min depending on resolution\n\n**Capabilities:**\n- Up to 20 seconds at up to 2K resolution\n- Simultaneous audio-visual generation (Dual-Branch Diffusion Transformer)\n- Text-to-video and image-to-video\n- Up to 12 reference files for multimodal input\n- OpenAI-compatible API structure\n\n**Ad creative use cases:**\n- High-volume short video ad production at low cost\n- Video ads with synchronized voiceover and sound effects in one pass\n- Multi-reference generation (feed product images, brand assets, style references)\n- Rapid iteration on video ad concepts\n\n**Docs:** [Seedance](https://seed.bytedance.com/en/seedance2_0)\n\n---\n\n### Higgsfield\n\nFull-stack video creation platform with cinematic camera controls.\n\n**Best for:** Social video ads, cinematic style, mobile-first content\n**Platform:** [higgsfield.ai](https://higgsfield.ai/)\n\n**Capabilities:**\n- 50+ professional camera movements (zooms, pans, FPV drone shots)\n- Image-to-video animation\n- Built-in editing, transitions, and keyframing\n- All-in-one workflow: image gen, animation, editing\n\n**Ad creative use cases:**\n- Social media video ads with cinematic feel\n- Animate product images into dynamic video\n- Create multiple video variations with different camera styles\n- Quick-turn video content for social campaigns\n\n---\n\n### Video Tool Comparison\n\n| Tool | Max Length | Audio | Resolution | API | Best For |\n|------|-----------|-------|------------|-----|----------|\n| **Veo 3.1** | 60 sec | Native | 1080p/4K | Gemini | Vertical social video |\n| **Kling 2.6** | 3 min | Native | 1080p | Third-party | Longer cinematic |\n| **Runway Gen-4** | 10 sec | No | 1080p | Official | Controlled, consistent |\n| **Seedance 2.0** | 20 sec | Native | 2K | Official + third-party | Affordable high-volume |\n| **Higgsfield** | Varies | Yes | 1080p | Web-based | Social, mobile-first |\n\n---\n\n## Voice & Audio Generation\n\nFor layering realistic voiceovers onto video ads, adding narration to product demos, or generating audio for Remotion-rendered videos. These tools turn ad scripts into natural-sounding voice tracks.\n\n### When to Use Voice Tools\n\nMany video generators (Veo, Kling, Seedance) now include native audio. Use standalone voice tools when you need:\n\n- **Voiceover on silent video** — Runway Gen-4 and Remotion produce silent output\n- **Brand voice consistency** — Clone a specific voice for all ads\n- **Multi-language versions** — Same ad script in 20+ languages\n- **Script iteration** — Re-record voiceover without reshooting video\n- **Precise control** — Exact timing, emotion, and pacing\n\n---\n\n### ElevenLabs\n\nThe market leader in realistic voice generation and voice cloning.\n\n**Best for:** Most natural-sounding voiceovers, brand voice cloning, multilingual\n**API:** REST API with streaming support\n**Pricing:** ~$0.12-0.30 per 1,000 characters depending on plan; starts at $5/month\n\n**Capabilities:**\n- 29+ languages with natural accent and intonation\n- Voice cloning from short audio clips (instant) or longer recordings (professional)\n- Emotion and style control\n- Streaming for real-time generation\n- Voice library with hundreds of pre-built voices\n\n**Ad creative use cases:**\n- Generate voiceover tracks for video ads\n- Clone your brand spokesperson's voice for all ad variations\n- Produce the same ad in 10+ languages from one script\n- A/B test different voice styles (authoritative vs. friendly vs. urgent)\n\n**API example:**\n```bash\ncurl -X POST \"https://api.elevenlabs.io/v1/text-to-speech/{voice_id}\" \\\n -H \"xi-api-key: $ELEVENLABS_API_KEY\" \\\n -H \"Content-Type: application/json\" \\\n -d '{\n \"text\": \"Stop wasting hours on manual reporting. Try DataFlow free for 14 days.\",\n \"model_id\": \"eleven_multilingual_v2\",\n \"voice_settings\": {\"stability\": 0.5, \"similarity_boost\": 0.75}\n }' --output voiceover.mp3\n```\n\n**Docs:** [ElevenLabs API](https://elevenlabs.io/docs/api-reference/text-to-speech)\n\n---\n\n### OpenAI TTS\n\nSimple, affordable text-to-speech built into the OpenAI API.\n\n**Best for:** Quick voiceovers, cost-effective at scale, simple integration\n**API:** OpenAI API (same SDK as GPT/DALL-E)\n**Pricing:** $15/million chars (standard), $30/million chars (HD); ~$0.015/min with gpt-4o-mini-tts\n\n**Capabilities:**\n- 13 built-in voices (no custom cloning)\n- Multiple languages\n- Real-time streaming\n- HD quality option\n- Simple API — same SDK you already use for GPT\n\n**Ad creative use cases:**\n- Fast, cheap voiceover for draft/test ad versions\n- High-volume narration at low cost\n- Prototype ad audio before investing in premium voice\n\n**Docs:** [OpenAI TTS](https://platform.openai.com/docs/guides/text-to-speech)\n\n---\n\n### Cartesia Sonic\n\nUltra-low latency voice generation built for real-time applications.\n\n**Best for:** Real-time voice, lowest latency, emotional expressiveness\n**API:** REST + WebSocket streaming\n**Pricing:** Starts at $5/month; pay-as-you-go from $0.03/min\n\n**Capabilities:**\n- 40ms time-to-first-audio (fastest in class)\n- 15+ languages\n- Nonverbal expressiveness: laughter, breathing, emotional inflections\n- Sonic Turbo for even lower latency\n- Streaming API for real-time generation\n\n**Ad creative use cases:**\n- Real-time ad preview during creative iteration\n- Interactive demo videos with dynamic narration\n- Ads requiring natural laughter, sighs, or emotional reactions\n\n**Docs:** [Cartesia Sonic](https://docs.cartesia.ai/build-with-cartesia/tts-models/latest)\n\n---\n\n### Voicebox (Open Source)\n\nFree, local-first voice synthesis studio powered by Qwen3-TTS. The open-source alternative to ElevenLabs.\n\n**Best for:** Free voice cloning, local/private generation, zero-cost batch production\n**API:** Local REST API at `http://localhost:8000`\n**Pricing:** Free (MIT license). Runs entirely on your machine.\n**Stack:** Tauri (Rust) + React + FastAPI (Python)\n\n**Capabilities:**\n- Voice cloning from short audio samples via Qwen3-TTS\n- Multi-language support (English, Chinese, more planned)\n- Multi-track timeline editor for composing conversations\n- 4-5x faster inference on Apple Silicon via MLX Metal acceleration\n- Local REST API for programmatic generation\n- No cloud dependency — all processing on-device\n\n**Ad creative use cases:**\n- Free voice cloning for brand spokesperson across all ad variations\n- Batch generate voiceovers without per-character costs\n- Private/local generation when ad content is sensitive or pre-launch\n- Prototype voice variations before committing to a paid service\n\n**API example:**\n```bash\ncurl -X POST http://localhost:8000/generate \\\n -H \"Content-Type: application/json\" \\\n -d '{\"text\": \"Stop wasting hours on manual reporting.\", \"profile_id\": \"abc123\", \"language\": \"en\"}'\n```\n\n**Install:** Desktop apps for macOS and Windows at [voicebox.sh](https://voicebox.sh), or build from source:\n```bash\ngit clone https://github.com/jamiepine/voicebox.git\ncd voicebox && make setup && make dev\n```\n\n**Docs:** [GitHub](https://github.com/jamiepine/voicebox)\n\n---\n\n### Other Voice Tools\n\n| Tool | Best For | Differentiator | API |\n|------|----------|---------------|-----|\n| **PlayHT** | Large voice library, low latency | 900+ voices, <300ms latency, ultra-realistic | [play.ht](https://play.ht/) |\n| **Resemble AI** | Enterprise voice cloning | On-premise deployment, real-time speech-to-speech | [resemble.ai](https://www.resemble.ai/) |\n| **WellSaid Labs** | Ethical, commercial-safe voices | Voices from compensated actors, safe for commercial use | [wellsaid.io](https://www.wellsaid.io/) |\n| **Fish Audio** | Budget-friendly, emotion control | ~50-70% cheaper than ElevenLabs, emotion tags | [fish.audio](https://fish.audio/) |\n| **Murf AI** | Non-technical teams | Browser-based studio, 200+ voices | [murf.ai](https://murf.ai/) |\n| **Google Cloud TTS** | Google ecosystem, scale | 220+ voices, 40+ languages, enterprise SLAs | [Google TTS](https://cloud.google.com/text-to-speech) |\n| **Amazon Polly** | AWS ecosystem, cost | Neural voices, SSML control, cheap at volume | [Amazon Polly](https://aws.amazon.com/polly/) |\n\n---\n\n### Voice Tool Comparison\n\n| Tool | Quality | Cloning | Languages | Latency | Price/1K chars |\n|------|---------|---------|-----------|---------|----------------|\n| **ElevenLabs** | Best | Yes (instant + pro) | 29+ | ~200ms | $0.12-0.30 |\n| **OpenAI TTS** | Good | No | 13+ | ~300ms | $0.015-0.030 |\n| **Cartesia Sonic** | Very good | No | 15+ | ~40ms | ~$0.03/min |\n| **PlayHT** | Very good | Yes | 140+ | <300ms | ~$0.10-0.20 |\n| **Fish Audio** | Good | Yes | 13+ | ~200ms | ~$0.05-0.10 |\n| **WellSaid** | Very good | No (actor voices) | English | ~300ms | Custom pricing |\n| **Voicebox** | Good | Yes (local) | 2+ | Local | Free (open source) |\n\n### Choosing a Voice Tool\n\n```\nNeed voiceover for ads?\n├── Need to clone a specific brand voice?\n│ ├── Best quality → ElevenLabs\n│ ├── Enterprise/on-premise → Resemble AI\n│ └── Budget-friendly → Fish Audio, PlayHT\n├── Need multilingual (same ad, many languages)?\n│ ├── Most languages → PlayHT (140+)\n│ └── Best quality → ElevenLabs (29+)\n├── Need free / open source / local?\n│ └── Voicebox (MIT, runs on your machine)\n├── Need cheap, fast, good-enough?\n│ └── OpenAI TTS ($0.015/min)\n├── Need commercially-safe licensing?\n│ └── WellSaid Labs (actor-compensated voices)\n└── Need real-time/interactive?\n └── Cartesia Sonic (40ms TTFA)\n```\n\n### Workflow: Voice + Video\n\n```\n1. Write ad script (use ad-creative skill for copy)\n2. Generate voiceover with ElevenLabs/OpenAI TTS\n3. Generate or render video:\n a. Silent video from Runway/Remotion → layer voice track\n b. Or use Veo/Seedance with native audio (skip separate VO)\n4. Combine with ffmpeg if layering separately:\n ffmpeg -i video.mp4 -i voiceover.mp3 -c:v copy -c:a aac output.mp4\n5. Generate variations (different scripts, voices, or languages)\n```\n\n---\n\n## Code-Based Video: Remotion\n\nFor templated, data-driven video ads at scale, Remotion is the best option. Unlike AI video generators that produce unique video from prompts, Remotion uses React code to render deterministic, brand-perfect video from templates and data.\n\n**Best for:** Templated ad variations, personalized video, brand-consistent production\n**Stack:** React + TypeScript\n**Pricing:** Free for individuals/small teams; commercial license required for 4+ employees\n**Docs:** [remotion.dev](https://www.remotion.dev/)\n\n### Why Remotion for Ads\n\n| AI Video Generators | Remotion |\n|---------------------|----------|\n| Unique output each time | Deterministic, pixel-perfect |\n| Prompt-based, less control | Full code control over every frame |\n| Hard to match brand exactly | Exact brand colors, fonts, spacing |\n| One-at-a-time generation | Batch render hundreds from data |\n| No dynamic data insertion | Personalize with names, prices, stats |\n\n### Ad Creative Use Cases\n\n**1. Dynamic product ads**\nFeed a JSON array of products and render a unique video ad for each:\n```tsx\n// Simplified Remotion component for product ads\nexport const ProductAd: React.FC<{\n productName: string;\n price: string;\n imageUrl: string;\n tagline: string;\n}> = ({productName, price, imageUrl, tagline}) => {\n return (\n \n \n

{productName}

\n

{tagline}

\n
{price}
\n
Shop Now
\n \n );\n};\n```\n\n**2. A/B test video variations**\nRender the same template with different headlines, CTAs, or color schemes:\n```tsx\nconst variations = [\n {headline: \"Save 50% Today\", cta: \"Get the Deal\", theme: \"urgent\"},\n {headline: \"Join 10K+ Teams\", cta: \"Start Free\", theme: \"social-proof\"},\n {headline: \"Built for Speed\", cta: \"Try It Now\", theme: \"benefit\"},\n];\n// Render all variations programmatically\n```\n\n**3. Personalized outreach videos**\nGenerate videos addressing prospects by name for cold outreach or sales.\n\n**4. Social ad batch production**\nRender the same content across different aspect ratios:\n- 1:1 for feed\n- 9:16 for Stories/Reels\n- 16:9 for YouTube\n\n### Remotion Workflow for Ad Creative\n\n```\n1. Design template in React (or use AI to generate the component)\n2. Define data schema (products, headlines, CTAs, images)\n3. Feed data array into template\n4. Batch render all variations\n5. Upload to ad platform\n```\n\n### Getting Started\n\n```bash\n# Create a new Remotion project\nnpx create-video@latest\n\n# Render a single video\nnpx remotion render src/index.ts MyComposition out/video.mp4\n\n# Batch render from data\nnpx remotion render src/index.ts MyComposition --props='{\"data\": [...]}'\n```\n\n---\n\n## Choosing the Right Tool\n\n### Decision Tree\n\n```\nNeed video ads?\n├── Templated, data-driven (same structure, different data)\n│ └── Use Remotion\n├── Unique creative from prompts (exploratory)\n│ ├── Need dialogue/voiceover? → Veo 3.1, Kling 2.6, Seedance 2.0\n│ ├── Need consistency across scenes? → Runway Gen-4\n│ ├── Need vertical social video? → Veo 3.1 (native 9:16)\n│ ├── Need high volume at low cost? → Seedance 2.0\n│ └── Need cinematic camera work? → Higgsfield, Kling\n└── Both → Use AI gen for hero creative, Remotion for variations\n\nNeed image ads?\n├── Need text/headlines in image? → Ideogram\n├── Need product consistency across variations? → Flux (multi-ref)\n├── Need quick iterations on existing images? → Nano Banana Pro\n├── Need highest visual quality? → Flux Pro, Midjourney\n└── Need high volume at low cost? → Flux Klein, Nano Banana\n```\n\n### Cost Comparison for 100 Ad Variations\n\n| Approach | Tool | Approximate Cost |\n|----------|------|-----------------|\n| 100 static images | Nano Banana Pro | ~$4-24 |\n| 100 static images | Flux Dev | ~$1-2 |\n| 100 static images | Ideogram API | ~$6 |\n| 100 × 15-sec videos | Veo 3.1 Fast | ~$225 |\n| 100 × 15-sec videos | Remotion (templated) | ~$0 (self-hosted render) |\n| 10 hero videos + 90 templated | Veo + Remotion | ~$22 + render time |\n\n### Recommended Workflow for Scaled Ad Production\n\n1. **Generate hero creative** with AI (Nano Banana, Flux, Veo) — high-quality, exploratory\n2. **Build templates** in Remotion based on winning creative patterns\n3. **Batch produce variations** with Remotion using data (products, headlines, CTAs)\n4. **Iterate** — use AI tools for new angles, Remotion for scale\n\nThis hybrid approach gives you the creative exploration of AI generators and the consistency and scale of code-based rendering.\n\n---\n\n## Platform-Specific Image Specs\n\nWhen generating images for ads, request the correct dimensions:\n\n| Platform | Placement | Aspect Ratio | Recommended Size |\n|----------|-----------|-------------|-----------------|\n| Meta Feed | Single image | 1:1 | 1080x1080 |\n| Meta Stories/Reels | Vertical | 9:16 | 1080x1920 |\n| Meta Carousel | Square | 1:1 | 1080x1080 |\n| Google Display | Landscape | 1.91:1 | 1200x628 |\n| Google Display | Square | 1:1 | 1200x1200 |\n| LinkedIn Feed | Landscape | 1.91:1 | 1200x627 |\n| LinkedIn Feed | Square | 1:1 | 1200x1200 |\n| TikTok Feed | Vertical | 9:16 | 1080x1920 |\n| Twitter/X Feed | Landscape | 16:9 | 1200x675 |\n| Twitter/X Card | Landscape | 1.91:1 | 800x418 |\n\nInclude these dimensions in your generation prompts to avoid needing to crop or resize.\n","skills/ad-creative/references/hook-system.md":"# The Hook System\n\nThe first three seconds decide whether the rest of the ad exists. Hooks are the highest-leverage unit of paid creative work — and hook *diversity* is what earns incremental learning: distinct hooks reach distinct pockets of the audience, while near-identical openings mostly re-test what you already know about the same one. This reference is a complete system for generating, diagnosing, and iterating hooks — not a list of one-liners.\n\nUse it inside Mode 1/3 generation (hooks for new concepts), Mode 2 iteration (diagnosing why an ad underperforms), and the creative strategy loop in [creative-roadmap.md](creative-roadmap.md).\n\n---\n\n## A Hook Is Three Components, Not a Line\n\nIn video, the hook is the simultaneous combination of:\n\n| Component | What it is | Job |\n|---|---|---|\n| **Visual action** | What is literally happening on screen in seconds 0–3 | Stop the thumb |\n| **Spoken line** | The first words of VO or dialogue | Open the loop |\n| **Caption text** | On-screen header/overlay text | Anchor the claim for sound-off viewers |\n\n**The no-duplication rule:** the three components must complement, never repeat. If the VO says \"I stopped paying $200/mo for my gym\" while the caption reads \"I stopped paying $200/mo\" over a static talking head, two of the three slots are wasted. Strong hooks split the work — visual shows the cancellation email, VO says the line, caption names the alternative. When writing hooks, write all three columns explicitly; a hook spec with one column filled in is a third of a hook.\n\nStatic ads collapse this to two components (visual + headline) — the same rule applies: the headline must not caption the image.\n\n---\n\n## The Generation Pipeline\n\nWork top-down; hooks written without the upstream steps read like everyone else's ads.\n\n```\nSegment → Motivation → Format → Hook (three components)\n```\n\n1. **Segment** — which specific buyer this hook addresses. Not the whole ICP: a slice with a shared situation (from the Grounded Inputs corpus: reviews, comments, sales-call language). The narrower the segment, the sharper the hook.\n2. **Motivation** — the single pain, desire, or objection that moves this segment, in *their* words. Pull verbatim phrases from reviews and comments; the corpus language always outperforms marketing paraphrase.\n3. **Format** — the delivery vehicle: street interview, POV selfie, screen recording, unboxing, side-by-side demo, text-on-screen static, founder-to-camera, reaction stitch. Pick the format *before* writing the line — the same motivation reads completely differently as a street-interview answer vs. a confession-to-camera.\n4. **Hook** — now write the three components for this segment × motivation × format cell.\n\n**Output as a hook matrix** so coverage is visible:\n\n```\n| # | Segment | Motivation (verbatim source) | Format | Visual action | Spoken line | Caption |\n```\n\nGenerate across the matrix, not down a single column — ten hooks for ten segment×motivation cells beat thirty rewordings of one cell. This is the same angle-diversity principle as the static template library: matrix diversity is audience diversity.\n\n---\n\n## Hook Opening Moves\n\nA menu of proven opening structures. Cycle through them like the static templates — don't cluster on favorites:\n\n| Move | Shape | Watch out |\n|---|---|---|\n| **Curiosity gap** | Withhold the noun: \"The ingredient behind most 3pm crashes\" | Must pay off within the ad or it's clickbait that poisons CVR |\n| **Bold claim** | A specific, falsifiable statement: \"This replaced my entire morning routine\" | Needs substantiation on screen or in the on-ramp |\n| **First-person confession** | \"I was doing [common thing] completely wrong\" | Reads fake without lived-in detail |\n| **Contrast / before-after** | Two states shown or named in the first beat | The transformation must be visually honest — see compliance notes in SKILL.md |\n| **Relatability / POV** | Mirror a hyper-specific situation: \"POV: it's 3pm and you're on your fourth coffee\" | Specificity is the entire mechanic; generic POV is invisible |\n| **Question** | Ask the exact question the buyer types into search or ChatGPT | Use their phrasing verbatim from the corpus |\n| **Countdown / gamified** | A timer or on-screen challenge that promises a payoff at the end | Payoff must exist; hold-rate collapses on cheats |\n| **Proof-first** | Lead with the receipt — the result screenshot, the stat, the demo money-shot | Strongest when the proof brags by itself |\n| **Choice** | Put two options in front of the viewer: \"Sandler or MEDDIC for a $50K deal?\" | Both options must be real ones the audience weighs; viewers who know both stay to see who wins, viewers who know neither stay to learn |\n| **Old vs. new** | \"We used to take elixirs. Now everyone drinks prebiotic soda. Which is better for your gut?\" | Answer the question from expertise; the hook only works if the payoff takes a side |\n| **Us vs. them** | \"Most [role]s do X. The ones who [result] do Y.\" | Name a behavior the audience will recognize as what \"most\" people really do |\n| **False assumption** | Challenge something the buyer believes they need: \"You can cancel ClickFunnels\" | Draws the strongest reactions, including angry ones; keep the body fair to the named product, and get legal review before naming a competitor in paid ads |\n| **Declaration, then question** | State what the audience already agrees with, then doubt it: \"You need to spend $20K a month recruiting on Indeed. Or do you?\" | The doubt needs a real answer in the body; still being tested by the practitioners who shared it (October 2026) |\n\nThe last five are one family: **contrast**. People understand a thing by comparing it with something else, so a hook that sets up two states, two options, or two groups gives the viewer a question to stay for. That is also why AI copy leans on \"it's not X, it's Y.\" The difference is that a good contrast hook compares two things the audience actually weighs, while the AI tell negates a claim nobody made (see the copywriting skill's AI-tells reference). Contrast hooks work best for educational content and decisions; for impulse products, lead with proof or POV instead.\n\n---\n\n## The Diagnostic Funnel\n\nEach metric in the delivery funnel isolates a different component. When an ad underperforms, read the funnel to find *which part* to fix instead of scrapping the whole ad:\n\n| Stage | Metric | If it's weak, the problem is | Fix |\n|---|---|---|---|\n| Stop | Thumbstop / 3-sec view rate | **Visual action** (and caption) | New visual opening; same everything else |\n| Stay | Hold rate (3s → 15s / 50% view) | **The on-ramp** — what follows the hook | Rework seconds 3–15, not the hook |\n| Click | CTR | Desire/offer clarity mid-ad | Sharpen the promise, CTA, or proof |\n| Convert | CVR post-click | Congruence — the page doesn't continue the ad | Fix the landing page or the claim, per **cro** |\n\nTwo rules this table enforces:\n\n- **A great thumbstop is not a great ad.** A clickbait visual that attracts the wrong viewers shows up as high thumbstop + collapsed hold/CVR. Read the whole funnel before declaring a winning hook.\n- **One component per iteration.** Change the visual OR the on-ramp OR the offer framing per test cycle — matching the one-variable rule in Common Mistakes.\n\n---\n\n## The On-Ramp Rule\n\nThe on-ramp is seconds ~3–15: the bridge from hook to body. **A good on-ramp logically extends the hook's premise; a bad one pivots to a product pitch that abandons it.** If the hook promises \"what actually causes this,\" the next beat must start explaining the cause — not introduce the brand story.\n\nCorollary: **every hook test is also an on-ramp test.** Swapping a new hook onto an existing ad body usually breaks the premise-bridge; when testing hooks, re-write the on-ramp to match each one. Hold rate is the on-ramp's metric — diagnose it separately from thumbstop.\n\n---\n\n## Fidelity Laddering\n\nMatch production cost to evidence strength (production tiers are defined in [creative-roadmap.md](creative-roadmap.md)):\n\n- **Hunches ship low-fidelity within a day or two:** statics, text-on-screen video, voiceover-over-b-roll, remixes of existing footage. The goal is a cheap signal on the *angle*, not a polished ad.\n- **Validated angles earn high-fidelity:** creator shoots, street interviews, staged demos. Only spend production budget on hooks whose low-fi version already showed a funnel signal (even a single-metric win — a hold-rate spike on an ugly static is evidence).\n\nTesting a hunch with an expensive shoot and testing a proven angle with a throwaway static are both mistakes — the ladder runs in one direction.\n\n---\n\n## Grounding Rules (inherited, non-negotiable)\n\nHooks inherit every grounding rule from SKILL.md: every hook cites the corpus source its motivation came from; no invented claims, stats, or testimonials; verbatim customer language over paraphrase. Additionally, mine **organic content in the niche** (top-performing TikToks/Reels/posts, via the **scraping** skill or the social listening tooling in **social**) for the audience's actual vocabulary — the words the niche uses (\"GLP-1\" vs. the clinical term, the slang for the pain) belong in the caption and spoken line. Organic mining is language research, not copying: take the vocabulary and the visual conventions, never a creator's specific creative.\n\n---\n\n## Common Failure Modes\n\n- **Thirty rewordings of one cell** — variation without matrix coverage; diversity of segment×motivation is the point\n- **Components duplicating each other** — three slots saying one thing\n- **Hook tested, on-ramp inherited** — premise-bridge broken, hold rate blamed on the hook\n- **Funnel read stops at thumbstop** — clickbait winners scale into CVR craters\n- **Polished hunches** — high-fidelity production spent on unvalidated angles\n- **Marketing-voice captions** — the corpus and the niche's organic content define the vocabulary; \"revolutionary formula\" appears in neither\n","skills/ad-creative/references/imessage-video-ads.md":"# iOS-Native Reveal Video Ads (iMessage, ChatGPT, Apple Notes, AirDrop)\n\nA family of 9:16 social-native video formats that recreate a familiar iOS surface in real time and let the brand emerge inside it. The flagship is the **iMessage chat reveal** — someone sends a screenshot of a result or product, a friend reacts and asks what it is, and the conversation reveals the brand, usually with a promo code. Message bubbles pop in over ~15–22 seconds with authentic send/receive sounds, then a static brand end card lands the CTA. The same architecture powers **ChatGPT reveals**, **Apple Notes reveals**, and **AirDrop reveals** — covered in [Other iOS-Native Reveal Surfaces](#other-ios-native-reveal-surfaces) below.\n\nThe format works because it borrows the most-read UI on earth. A chat thread is a familiar, high-attention dramatization — it mirrors how real recommendations happen, so the viewer leans in instead of scrolling past. The CTA arrives conversationally (\"use code FREEPACK\") instead of as a hard sell, which keeps the ad-skip reflex from firing until the pitch has already landed. Run it only as a clearly labeled paid placement (Meta's \"Sponsored\" tag does the disclosure work); never seed it organically as if it were a real leaked conversation.\n\nCredit: this reference distills the format popularized by Shiv Sakhuja and the Gooseworks team ([@shivsakhuja](https://x.com/shivsakhuja), [gooseworks-ai/gooseworks-ads-skills](https://github.com/gooseworks-ai/gooseworks-ads-skills)), who report the format performing strongly on Meta.\n\n---\n\n## When to Use This Format\n\n**Good fit:**\n- Reaction/discovery ads where the punchline is the recipient's curiosity (\"wait, what app is that?\")\n- Promo-code offers — the conversational delivery feels far less ad-like than a code on a slate\n- Products with a screenshot-able result: a number, a dashboard, a receipt, a before/after\n- UGC-style angles when you don't have UGC creators on tap\n\n**Poor fit:**\n- Considered B2B purchases where a casual text exchange undercuts credibility\n- Products with nothing visual or numeric to screenshot (fix the hook first, not the format)\n- Brands whose compliance review can't approve dramatized conversations (regulated industries — check first)\n\n**Platform fit:** Built for Meta Reels/Stories placements (9:16, 1080×1920) with a 1:1 center-crop variant for feed. Works on TikTok and YouTube Shorts with the same master file.\n\n---\n\n## Compliance and Grounding\n\nThis is a **dramatization** — a scripted conversation, not a real one. That's a standard, legitimate ad device, but two rules keep it honest and on the right side of FTC guidance:\n\n1. **Every claim in the thread must be true of the product.** The race time, the savings math, the \"5 minutes a day\" — ground each one in a real customer result, review, or verifiable product fact, exactly as the Grounded Inputs rules in SKILL.md require. The conversation is fictional; the facts inside it can't be.\n2. **Don't present the thread as a real testimonial.** No real customer names, no \"this is an actual text from a customer\" framing, no fabricated endorsements. The format persuades through recognizability, not through pretending to be found footage.\n\nIf a claim needs a disclaimer on your landing page, it needs one on this ad too.\n\n---\n\n## Concept Angles\n\nMost iMessage ads fit one of six angles. Pick the angle before writing any copy — the most common failure mode (\"script is fine but the ad feels off\") is an angle mismatch, not bad lines. The strongest hooks share one of three traits: a specific number, a small act of self-trust, or a physically novel product mechanic.\n\n| Angle | The hook attachment | The reveal |\n|---|---|---|\n| **Result-as-screenshot** | A number that brags by itself — race time, app summary, dashboard stat | \"X minutes a day. that's it.\" |\n| **Setup flex** | A photo of your space — tiny apartment gym, race-kit corner, desk setup | \"this is the whole setup\" |\n| **Cancellation moment** | A confirmation receipt — gym cancellation email, \"subscription cancelled\" page | \"$X/mo → $Y/mo. do the math\" |\n| **Feature-as-punchline** | A short clip of the product mechanic in motion | The mechanic *is* the brand |\n| **Friend-asks-friend (inverse)** | The *peer* opens with the wow — \"how are you doing this 😭\" | *You* reply with the brand |\n| **Receipt-as-hook** | A mundane financial document — statement, App Store receipt | A small act of self-trust |\n\n---\n\n## Anatomy of the Ad\n\n```\n0:00 Hook attachment lands (the screenshot the whole chat is about)\n ↓ short reactions, 250–450ms apart (\"bro no way\" / \"wait is that real\")\n0:06 The question — \"what app is that??\"\n ↓ typing indicator … then the brand-name reply\n0:12 The pitch, in texting voice — one or two bubbles max\n0:15 The code — \"use FREEPACK, first pack's free\" (code renders link-underlined)\n0:17 Beat of silence, then the closer — \"bet\" / \"ok downloading\"\n0:18 300ms crossfade → static brand end card: logo, code, tagline (~3s)\n```\n\n**Script rules:**\n\n- **8–14 bubbles total.** Shorter reads thin; longer loses the scroll-past viewer.\n- **Write in real texting voice.** Lowercase, fragments, one emoji max per message, no marketing adjectives. Read it aloud as two friends — any bubble that sounds like ad copy gets cut.\n- **The brand appears once, late.** The thread is about the *result* until someone asks. Naming the brand in bubble two kills the reveal.\n- **Pacing has rhythm, not a metronome.** One-word reactions fire 250–450ms apart; sentence replies get 600–900ms of air after them; leave ~600ms of silence before the final reaction so it lands.\n- **Typing indicators go before sentence-length peer replies**, optional before short reactions. The indicator appearing is silent (see SFX rules below).\n- **The promo code goes inside a bubble**, styled with iOS's link-detection underline, *and* on the end card. Conversational delivery first, reinforcement second.\n\n---\n\n## Production Routes\n\nTwo ways to produce it, in order of control:\n\n### Route 1: Code-based pipeline (full control)\n\nThe architecture that produces a convincing result: render the chat as HTML/CSS mimicking the iMessage UI, drive the animation with a timeline script, record it headlessly with Playwright, and assemble audio + end card with ffmpeg.\n\n1. **Script as data.** Store the thread as JSON: participants (peer name, initials, avatar color), ordered messages (`from`, `text`, attachment paths, typing-indicator flags), theme, header. The script is reviewable and re-renderable without touching code.\n2. **Render the chat UI in HTML/CSS.** Dark theme reads most native. Two variants: full-bleed chat, or the chat inside an iPhone frame (status bar + Dynamic Island) over a brand-relevant background photo — the framed variant reads more native in-feed and is the better default.\n3. **Animate with a timeline, record in ONE continuous session.** All bubbles exist in the DOM but hidden (`display: none` — not `opacity: 0`, or the thread pre-allocates space and never \"grows\"). A driver script walks a timeline array revealing each bubble, driving the composer, and auto-scrolling. Never record scene-by-scene and concat — every page reload causes a visible micro-flicker.\n4. **Type the composer for every sent bubble.** The typed text must exactly equal the sent text (a mismatch reads fake on second watch). Pace ~12–15 chars/sec with ±30% per-character jitter so it feels like thumbs, not a script.\n5. **Record at native output resolution.** Set both the Playwright `viewport` *and* `recordVideo.size` to 1080×1920 — if you omit `recordVideo.size`, Playwright records a scaled-down video by default. Recording small and upscaling ships soft, blurry bubble text.\n6. **Layer audio with ffmpeg.** SFX cues computed deterministically from the same timeline that drove the recording, so sounds land exactly on bubble pops.\n7. **Stitch: chat → 300ms crossfade → static end card.** ffmpeg's `xfade` requires both inputs to match in resolution, pixel format, and frame rate — render the end card to a fixed-frame MP4 at the same specs as the chat recording before fading. Export the 9:16 master plus a 1:1 center crop.\n\n### Route 2: Remotion (templated scale)\n\nOnce a winning script structure emerges, rebuild it as a Remotion composition (see [generative-tools.md](generative-tools.md)) with the thread JSON as props. Then variations — new hooks, new codes, new personas — are data changes, not re-productions. Right move at the \"we're testing 10 script variants a week\" stage, not for the first ad.\n\n---\n\n## Craft Rules (the details that sell the illusion)\n\nThese are the difference between \"feels like a real chat\" and \"feels like a mockup\":\n\n- **The real send/receive sounds, never generic notification sounds.** The iMessage feel is mostly the audio. BigSoundBank hosts recordings of Apple's message sounds under CC0: send whoosh (`bigsoundbank.com/UPLOAD/mp3/1313.mp3`, ~0.5s) and receive tritone (`bigsoundbank.com/UPLOAD/mp3/1111.mp3` — trim to ~1.4s with a 400ms fade). Normalize loud (≈ -9 LUFS) so they cut through the music. Note the recordings being CC0 doesn't mean Apple has licensed its sound marks or UI trade dress — this is standard practice in the format, but regulated brands and risk-averse legal teams should review the iMessage mimicry as a whole; a generic chat-app skin (neutral bubbles, non-Apple sounds) is the fallback that keeps the mechanic.\n- **No sound on the typing indicator.** iOS is silent when someone starts typing. Play the receive sound only when the actual bubble replaces the dots. This is the single most common tell.\n- **Music bed: quiet lofi/hip-hop instrumental.** ~30% volume, highpass around 60Hz to clear room for the SFX, fade out ~1.5s before the code reveal so the CTA lands in relative silence.\n- **Static end card — no zoom, no Ken Burns drift.** The brand slate must land hard; a drifting end card reads as filler.\n- **Real brand logo SVG on the end card, never CSS-styled text.** Font-approximated wordmarks look amateur even when close. Pull the official SVG from the brand's press kit, Wikimedia, or brandfetch.com.\n- **Hook screenshots: mimic the real app's UI, don't AI-generate it.** AI-generated app UIs ship garbled chrome that reads as slop. Build a small HTML page copying the actual app's brand colors, typography, and layout conventions (the Strava-orange strip, the \"Public · 2h ago\" timestamp) and screenshot it. Reserve AI image generation for *photographic* hooks — a beach photo, a lifestyle shot, the framed variant's background.\n- **Audio mixing gotcha:** ffmpeg's `amix` divides volume by input count by default — pass `normalize=0` or the whole mix comes out mysteriously quiet. Then run the mix through a limiter with the ceiling just under full scale (e.g. `alimiter=limit=0.95`, ≈ -0.4 dB) so it's loud without clipping.\n\n---\n\n## Quality Checklist\n\nBefore shipping:\n\n- [ ] Every factual claim in the thread traces to a real review, result, or product fact (Grounded Inputs)\n- [ ] Script reads as real texting voice when read aloud — no marketing adjectives in bubbles\n- [ ] Brand name appears only after the peer asks\n- [ ] No sound on any typing indicator; receive SFX fires when the text bubble lands\n- [ ] SFX land exactly on bubble pops (spot-check first and last)\n- [ ] Every sent bubble had a full composer drive; typed text equals sent text\n- [ ] No micro-flicker anywhere in the chat — the only cut is chat → end card (300ms crossfade)\n- [ ] Promo code is link-underlined in its bubble and repeated on the end card\n- [ ] End card is static with the real logo SVG\n- [ ] Master is native 1080×1920; 1:1 variant is a crop, not a squeeze\n- [ ] Final bubble gets ~600–800ms of air before the crossfade\n- [ ] Audio is limited just under full scale (no clipping); music never fights the SFX\n\n---\n\n## Iterating the Format\n\nTreat the thread as the variable and the pipeline as fixed. Test in this order — hook first, everything else after:\n\n1. **Hook attachment** — the screenshot is the thumbnail and the first 2 seconds; it decides the scroll-stop\n2. **Angle** — result-flex vs. cancellation vs. inverse changes who the viewer identifies with\n3. **Code reveal phrasing** — \"first pack's free with FREEPACK\" vs. \"FREEPACK gets you one free\"\n4. **Peer persona** — name, avatar, and texting style shift the perceived audience\n5. **Length** — try a 12-bubble and an 8-bubble cut of the same script\n\nThe same architecture extends to further surfaces too — WhatsApp, Slack, a search box — same timeline-driven recording, different UI shell.\n\n---\n\n## Other iOS-Native Reveal Surfaces\n\nEverything above about production (UI mockup → timeline-driven continuous recording → deterministic SFX cues → static end card), grounding, and disclosure carries over unchanged. What changes per surface is the *persuasion mechanic* and a handful of craft details.\n\n| Surface | Persuasion mechanic | Reach for it when |\n|---|---|---|\n| **iMessage** | A friend's recommendation — social proof through dialogue | The product is discovered through results people share (\"what app is that?\") |\n| **ChatGPT** | An authoritative answer to the viewer's own question | The problem is question-shaped — something people would literally type into ChatGPT |\n| **Apple Notes** | A private confession made public — first-person, no dialogue | The angle is transformation or realization (\"things nobody told me about 45\") |\n| **AirDrop** | A spontaneous peer share — \"someone nearby thought this was worth sending you *right now*,\" with a built-in accept/decline decision | The product is something people pass to each other (a deal, a link, a find, a file) and the accept-tap can *be* the reveal |\n\nThe strongest signal for choosing: which of these surfaces already fills your audience's day. Recommendation products want iMessage; advice-seeking problems want ChatGPT; identity/transformation stories want Notes; and anything people spontaneously pass to each other wants AirDrop.\n\n### ChatGPT Reveal\n\nThe viewer identifies with the *asker*. The typed question is the hook and must be the target customer's verbatim question — awkward phrasing and all (\"why is my stomach so bloated all of a sudden at 47?\"). The streaming answer names the problem's real mechanism, then the solution category; the brand lands in the answer's recommendation or in a typed follow-up (\"what's the best one?\").\n\n**Craft details:**\n- **Stream the answer in word chunks**, not character-by-character (that's typing, not generation) and not whole paragraphs at once. A subtle tick underneath the stream and a clean stop when the response completes; no iMessage tritones anywhere.\n- **Type the question like thumbs, stream the answer like a model.** Two distinct rhythms — the contrast is what reads as \"real ChatGPT.\"\n- **Keep the answer scannable:** short paragraphs, a bolded phrase or a short list, exactly the way ChatGPT actually formats. A wall of text breaks the illusion and loses the viewer.\n- OpenAI's interface is their trade dress — same legal-review posture as the Apple UI mimicry note above, with a generic \"AI assistant\" skin as the fallback.\n\n**Compliance — stricter here than anywhere else in this family.** The \"answer\" is your ad copy wearing a lab coat: an authority costume. Every claim in it needs the same substantiation as a claim in your own voice, and the format's borrowed authority raises the bar, not lowers it. Do not put health, medical, or financial advice in a fabricated AI answer without legal review — that's the highest-risk version of this format. And never present the exchange as a real, unprompted ChatGPT output endorsing your product; it's a dramatization, same as the iMessage thread.\n\n### Apple Notes Reveal\n\nA different genre from the chat formats: **confession, not conversation.** The viewer watches someone type a private note — a list of realizations, a \"things I wish I knew\" entry — with the keyboard visible. The note's title is the hook and does the job slide 1 does in a carousel (\"Things nobody told me about 45.\"). The product appears as one item in the list, named the way a person would actually write it to themselves — not the way a brand would.\n\n**Craft details:**\n- **Audio is keyboard taps only.** No chat SFX, no receive tones — a note has no other party. A quiet music bed still works underneath.\n- **Type at real thumb pace with jitter**, same as the iMessage composer rule. One typo-and-correction reads as human; several read as staged.\n- **Get the Notes chrome right:** title styled larger than body, the formatting bar above the keyboard, iOS-yellow accents. Same HTML-mimicry approach — and the same Apple trade-dress review note and generic-notes-app fallback — as everything else here.\n- **Fit the note to the frame.** Write short enough that the whole note fits without scrolling, or scroll once, deliberately, late.\n- **First person or it doesn't work.** The moment the note reads like ad copy (\"[Brand] changed everything!\"), the intimacy that makes the format convert is gone. The product mention should be the *least* enthusiastic line in the note.\n\nThe grounding rule hits differently here: the confession is a dramatization of a *composite, true* customer story — pull the realizations from real reviews and interviews (the Grounded Inputs corpus), and keep any numbers or outcomes to documented ones.\n\n### AirDrop Reveal\n\nThe one interaction-native format in the family: the hook is an **incoming AirDrop request**, and the **Accept tap is the reveal**. The viewer watches from the *receiver's* POV — a translucent AirDrop card slides up, \"[Sender] would like to share [preview],\" with a gray Decline and a blue Accept. The curiosity is structural (\"what is this and who's sending it?\") and the accept/decline choice is a built-in micro-conversion beat baked into iOS itself. Tapping Accept transfers the item — and *that's* where the product, the offer, or the result lands.\n\n**Craft details:**\n- **The preview thumbnail is the hook.** It's the one image on the AirDrop card before Accept, so it has to earn the tap — same job as the iMessage screenshot attachment. Make it the result, the product money-shot, or the offer.\n- **Cast the sender name like a real share.** \"Sarah's iPhone,\" \"Mom,\" \"Jordan's MacBook\" reads native; a brand name in the sender slot reads like an ad — save brand-as-sender for the reveal, not the incoming card.\n- **The transfer progress ring is the signature motion — don't skip it.** Incoming card → a beat of hesitation (\"accept?\") → the Accept tap → the circular progress fills → the item lands + end card. That progress-ring beat is what makes it read as a real AirDrop and not a cut.\n- **Audio is the AirDrop swoosh / received tone**, not the iMessage tritones. Same CC0-Apple-sounds sourcing and the same Apple trade-dress review note as the rest of the family, with a generic \"nearby share\" skin as the fallback.\n- **Keep it short and get the material right.** The card's blur/translucency and the gray Decline / blue Accept button pair are the recognizable cues; a flat opaque sheet breaks the illusion. The whole beat is faster than the chat formats — the interaction *is* the ad.\n- **Receiver POV by default; sender POV as the flex.** Receiving reads as discovery (\"someone sent me this\"); sending reads as a recommendation you're making (\"had to AirDrop this to the group\") — use sender POV when the angle is advocacy rather than discovery.\n\nGrounding is the same family rule: it's a dramatization of a share, not a claim that a real person actually AirDropped your product. Every claim on the transferred item is substantiated per the Grounded Inputs rules, and the exchange is never presented as a real, unprompted endorsement.\n","skills/ad-creative/references/meta-creative-formats.md":"# Meta Creative Format Taxonomy — Which Format to Make Next\n\nA prioritized S→F catalog of ~51 Meta ad creative formats, built as a **decision aid for \"which format do I make next,\"** not an encyclopedia. Use it to pick a format before you brief it, and to stop pouring hours into formats that structurally can't do the job you need.\n\nDistilled from Dara Denney's public tier list (10 yrs on Meta, teams that shipped ~20,000 creatives), re-expressed in this skill's voice — patterns credited, descriptions not copied.\n\n## The one question that ranks everything\n\nFor any format, ask: **is this a *unicorn scaler* or a *supporting cast member*?**\n\n- **Unicorn scaler** — punctures *cold, net-new* audiences and holds up as you scale spend. These are rare and worth disproportionate investment.\n- **Supporting cast** — converts people already in the mid/low funnel. Useful, necessary, but it will *not* open new audiences no matter how much you spend on it.\n\nThat distinction is the whole ranking. A format isn't \"bad\" for being supporting cast — it's bad only when you expect it to scale into cold audiences and it structurally can't. **Build a portfolio:** a few unicorn scalers doing the puncturing, a bench of supporting cast doing the converting.\n\n## Why creator-fronted formats top the list (Andromeda)\n\nMeta's **Andromeda algorithm is persona-based** — it targets *personas*, not just interests. Creator-fronted formats win because they reach a persona *natively*: through a creator that persona already follows and trusts. The seed audience for a partnership ad literally starts from the creator's own audience. That's why founder content, partnership ads, and authority ads dominate the top — the format is doing the targeting.\n\n**Practical signal to watch:** track rolling month-over-month *reach*. When it falls, you've saturated your current audience — deploy creator-fronted formats (especially partnership ads) to restore net-new reach.\n\n## Production complexity legend\n\n- **Low** — copy + one asset; you can make it today (statics, founder's letter, text-driven).\n- **Med** — needs a creator, a shoot, a script, or an edit (yapper, green-screen, VSL script).\n- **High** — multi-party, rights, or heavy production (celebrity, warehouse shoot, AI animation, press).\n\n---\n\n## S-tier — unicorn scalers (invest here first)\n\n| Format | Funnel role | Complexity | When to use / when to skip |\n|---|---|---|---|\n| **Founder content** | Cold scaler | Low–Med | The reliable *first* winner at any production level. Tell the story of *why* you built the brand — you auto-connect with same-problem buyers. **Use** early, when you have no proven creative yet. Rarely a skip. |\n| **Partnership ads** | Cold scaler | Med | **#1 investment priority.** \"Making or breaking brands on Meta right now\"; not running them is \"a butter knife to a gunfight.\" Best path to personas + net-new reach. **Use** always, and deploy when rolling reach drops. Skip only if you genuinely can't source creators. See #529. |\n| **VSL (video sales letter)** | Cold scaler | Med–High | Top-tier for anything that needs upfront **education** — health, wellness, fitness, complex mechanisms. **Use** when the buyer must understand *why it works* before buying. **Skip** for impulse/low-consideration products. Build the copywriting craft; the script is the ad. |\n\n**S-tier tactic:** when you contract creators for partnership ads, *also* have each shoot a few low-fi creator statics (how they'd post a Story for the brand). Builds a mini-funnel per creator for near-zero marginal cost.\n\n---\n\n## A-tier — scales up nicely\n\nCold-capable with the right inputs; the next tier to test once your S-tier is running.\n\n| Format | Funnel role | Complexity | When to use / when to skip |\n|---|---|---|---|\n| **Amateur investigation** | Cold scaler | Med | A creator \"investigates\" your product/niche (e.g. visiting competitors). Fresh, high-engagement. **Use** in categories where skepticism is the barrier. |\n| **Yapper ads** | Cold scaler | Med | Creator yaps to camera with personal storytelling. **High ceiling, hard to nail** — needs the *right* creator + script + setting. **Skip** if you can't cast well; a mediocre yapper flops. |\n| **David & Goliath** | Cold-capable | Low–Med | Position the brand as David vs. a big incumbent/obstacle; storytelling makes people root for you. **Use** when there's a clear villain (legacy category, bloated competitor). |\n| **Grid-style statics** | Cold-capable | Low | Multi-product / SKU / bundle grid. Easy to make, was a top performer at a 9-figure brand. **Lowest-hanging fruit to test** — make some this week. |\n| **Authority ads** | Cold scaler | Med | A doctor/dermatologist/expert fronts it. **Use** in hyper-competitive, trust-gated niches (supplements, beauty). Adds validation + creative diversity beyond UGC. |\n| **Green-screen commentary** | Cold-capable | Med | Creator composited over content, commenting. **Use** in apparel especially, with an educational angle. |\n| **Catalog / DPA** | Cold-capable | Low–Med | **Under-used truth:** not just retargeting — can run top-of-funnel/cold prospecting (DABA). Most brands leave this on the table. **Use** with a real catalog; currently a top performer for some accounts. |\n\n---\n\n## B-tier — solid supporting cast\n\nConvert mid-funnel reliably; occasionally sneak into the top rotation with great messaging. Don't expect them to open cold audiences. Most are **Low** complexity (statics) unless noted.\n\nTikTok love letter · Real short *(top-of-funnel support, Med)* · Callout ads · Before/after *(mid-funnel; watch claims)* · Progression *(mid-funnel)* · Tweet/Reddit statics *(great as the **first frame**; good in the $100k–250k spend range)* · Headline ads *(OG print-era; needs **amazing** messaging, pairs with callouts)* · Us-vs-them *(mid-funnel; sneaks into the top 8)* · Hot-girl IG stories *(mirror selfies / flat-lays)* · Creator low-fi statics *(the partnership tactic above)* · Objection-handling *(works fast, often top-15)* · Founder's letter static *(cranks during sales)* · Conversation ads *(Med; hard to execute)* · Educational infographics *(masquerades as content; under-used)* · Mood board *(apparel)* · Comment-reply · Challenging-your-beliefs *(Med; needs B-roll + known persona beliefs)* · Ugly / handwriting / post-it *(crush during sales periods)*\n\n---\n\n## C-tier — situational / operationally complex\n\nCan win in narrow conditions but cost more than they return for most accounts. Reach for these only when the specific condition applies.\n\nAI animation *(Pixar/claymation; High — hits net-new pockets initially, rarely holds long-term)* · Statistics ads *(luxury/retail + awareness/traffic objectives, **not** D2C ROI)* · Celebrity *(High; can crank or be a money pit)* · AI avatar *(has scaled **with** legal disclaimers, but phasing out as brands pick real creators)* · Warehouse *(High; great for sales, complex to shoot)* · Street interview *(often better to **fake/recreate** than capture live)* · Duet/reaction/stitch *(needs rights from the original creator)* · ASMR *(pet/beauty; needs specific ASMR creators)* · Regular UGC *(still works, but **general fatigue** on manufactured problem-solution VO + B-roll UGC)*\n\n---\n\n## D-tier — rarely moves the needle\n\nBreaking-news ads *(born to replace unreliable press)* · AI billboard *(overdone/cheesy; only lands with punchy/taboo language in supplements)* · GRWM / day-in-my-life *(organic-native; doesn't scale on paid unless the product fits a morning routine)*\n\n---\n\n## E-tier — mostly skip\n\nText-only *(usually executed with bland AI copy; exception: founder's letter during sales)* · Testimonial statics *(marketers execute them badly — only worth it with golden-nugget testimonials)* · Listicles *(worked a year or two ago, dead lately)* · Carousel *(juice rarely worth the squeeze — multiple assets, unknown payoff)*\n\n---\n\n## F-tier — don't bother\n\nExplicitly de-prioritized. These aren't just weak — they cost real time/rights and reliably underperform.\n\n- **Press ads** — a rights/permissions nightmare now (Vogue et al. will come after you). Was a champion format years ago; the ground shifted.\n- **Podcast ads** — a waste unless a **founder is on an actually well-known show**. Renting a studio or AI-generating a fake podcast clip doesn't pay off.\n- **Notes-app / UX fake-native ads** — everywhere on guru reels, but **they do not convert**. The familiar UI makes *everyone* stop, so they fail to qualify the right people and **confuse the algorithm**. Skip regardless of how tempting the \"native\" look is.\n\n---\n\n## Cross-cutting principles\n\n- **Portfolio, not silver bullet.** Only founder / partnership / authority / investigation / VSL / grid-static reliably scale cold. Everything else is a converter — staff both roles.\n- **Andromeda is persona-based** → creator-fronted formats win because the format *is* the targeting.\n- **Fake it when honest capture is painful** — street interviews and duet reactions can be recreated; don't wait for the perfect real moment.\n- **Fatigue is real** on over-taught formats (manufactured UGC, notes-app, AI billboards). **Freshness itself is an edge** — a novel-but-honest format out-punches a saturated \"best practice.\"\n\n---\n\n## Where the details live\n\nThis file is the **format map** — priority and selection. The *how-to-build* lives elsewhere:\n\n- **Static formats** (grid, us-vs-them, headline, callout, before/after, founder's letter, FAQ, tweet/Reddit, etc.) → structural templates with copy slots in [static-ad-templates.md](static-ad-templates.md).\n- **Video formats** (VSL, yapper, green-screen, UGC reaction, faceless/motion, iOS-native reveals) → the vertical-video production spec + creator-format library in [short-form-video-specs.md](short-form-video-specs.md), the motion-style pipeline in [motion-video-ads.md](motion-video-ads.md), and the iOS-native reveals in [imessage-video-ads.md](imessage-video-ads.md).\n- **Deciding which specific concepts to make** (evidence-ranked, account-state-aware) → the Creative Strategy Loop in [creative-roadmap.md](creative-roadmap.md).\n- **Kill/keep/scale math** once these are live → `ads` skill's [meta-decision-system.md](../../ads/references/meta-decision-system.md).\n\n*Tier list and the unicorn-vs-supporting-cast framing adapted from Dara Denney's \"I Ranked 51 Meta Ad Creative Types (Tier List)\"; yapper/investigation craft informed by Oren John. Patterns credited, descriptions re-expressed. Tiers reflect a point in time — Meta's algorithm and format fatigue shift; re-verify against current account data.*\n","skills/ad-creative/references/motion-video-ads.md":"# Motion-Style Video Ads (Faceless, Fully Generated)\n\n> Format popularized by Borja ([@borjafat](https://x.com/borjafat)) and the open `super-video-maker` motion-collage recipe by [Bomx](https://github.com/Bomx/super-video-maker-skill); this guide is an original re-expression of the method, extended with a multi-style library and production lessons from building and shipping it end-to-end.\n\nProduce a 15–45s faceless video ad or explainer from nothing but a concept: a styled\nposter still (image model) → brought to life with subtle motion (image-to-video model)\n→ narrated (TTS) → word-timed captions. No footage, no presenter, no editor. Cost per\nfinished video is roughly $3–6 in API calls; wall-clock ~15 minutes.\n\nThe format works because the *still* carries the idea (one literal, slightly surreal\nvisual per beat) and the *motion* only makes it breathe. Resist the urge to make the\nvideo do the storytelling — this is animated poster design, not filmmaking.\n\n## When to use\n\n- Concept/explainer ads: one idea made concrete (\"your CRM is a junk drawer\")\n- Top-of-funnel social video (9:16 Reels/Shorts/TikTok, 4:5 and 1:1 feed)\n- Brand-response hybrids where a distinctive owned style beats stock UGC\n- NOT for: demo/proof ads (screen recordings win), testimonial/UGC formats,\n anything requiring a real product shot as evidence\n\n## Pipeline (provider-agnostic)\n\n1. **Script** 3–6 beats, 20–45s of VO. One idea per beat. Calm and specific beats\n hype. End on a single CTA line.\n2. **Poster stills** — one per beat, using a *style formula* (below). Generate beat 1,\n approve it, then pass it as a reference image for every later beat so the set reads\n as one series. Fix garbled label text by regenerating with a shorter phrase.\n3. **Animate** each approved still with an image-to-video model (5–8s per beat).\n Motion belongs to the objects in the frame; the composition must not change.\n4. **VO + captions**: one continuous TTS take, transcribe with word timestamps\n (whisper), cut beats at sentence boundaries, burn 2–3-word caption groups.\n5. **Assemble**: concat beats trimmed to their VO spans (hold the last frame to pad),\n loudness-normalize to `I=-16:TP=-1.5:LRA=11`, export per-placement aspect.\n\n**Provider options** (any combination works; the recipe is model-agnostic):\n\n| Stage | One-key Gemini path | Alternatives |\n|---|---|---|\n| Stills | Nano Banana Pro (`gemini-3-pro-image-preview`) — excellent label typography | GPT-Image, Flux, Ideogram |\n| Motion | Veo 3.1 fast image-to-video (note: 1080p requires 8s clips) | Seedance 2.0 via fal.ai, Kling, Runway |\n| VO | Gemini TTS (calm voices: Charon/Kore) | ElevenLabs, OpenAI TTS |\n| Captions | whisper word timings + PIL/ASS burn-in | CapCut, platform auto-captions |\n\n## The style library\n\nFive proven looks. Each is a fill-in-the-slots prompt formula; keep ONE style per\ncampaign so the account builds a recognizable visual identity. All five animate well.\n\n### A. Screen-print collage (editorial, \"In a Nutshell\" docu energy)\n> Flat screen-print collage poster, single saturated `` background, subtle newsprint grain. Centerpiece: a black-and-white halftone cutout of ``, treated as a paper sticker with a thin white die-cut outline, slightly torn edges, and a soft drop shadow. Visible halftone dot texture, vintage editorial photo feel, grayscale subject. Accent cutouts: 2–4 flat shapes (cream circle sun, black zigzag, scattered dots). A torn-paper label near the bottom with the words \"`