Update to Marketing Swarm

{
  "name": "creative-genome-analysis",
  "description": "Use to analyze one or more ads or creative variants, identify the message and execution patterns associated with performance, and turn winners into testable creative hypotheses without copying them blindly.",
  "included_files": [
    {
      "relative_path": "agents/openai.yaml",
      "size_in_bytes": 302
    }
  ],
  "skill_md_contents": "---\nname: creative-genome-analysis\ndescription: Use to analyze one or more ads or creative variants, identify the message and execution patterns associated with performance, and turn winners into testable creative hypotheses without copying them blindly.\n---\n\n# Creative Genome Analysis\n\nBreak creative into analyzable components, then relate those components to audience response and performance evidence.\n\n## Creative representation\n\nFor each asset, capture only what is actually observable or provided:\n\n- hook: question, claim, contrast, demonstration, story, statistic, direct offer, other\n- audience/problem cue\n- promise or desired outcome\n- mechanism or reason-to-believe\n- proof: demonstration, testimonial, data, social proof, authority, guarantee, none\n- offer and price framing\n- CTA\n- visual opening / first-frame idea\n- product visibility\n- pacing and format\n- copy length and density\n- brand presence\n\nDo not force every ad into a fixed taxonomy when the format does not fit.\n\n## Performance linkage\n\nWhen creative-level data exists, compare elements against relevant metrics. Avoid declaring an element a winner because it appears in one winning ad.\n\nLook for:\n\n- repeated patterns across multiple strong creatives\n- interaction effects between message and format\n- top-of-funnel attention versus downstream conversion quality\n- changes after audience, placement, or spend shifts\n- sample size and exposure imbalance\n\nClassify findings as:\n\n- **observed pattern**\n- **plausible mechanism**\n- **untested hypothesis**\n\n## Build the next test\n\nTurn analysis into a controlled matrix. Prefer changing one major hypothesis dimension at a time unless the goal is exploratory concept testing.\n\nExamples of test dimensions:\n\n- hook angle\n- proof format\n- offer framing\n- demo versus testimonial\n- product-first versus problem-first opening\n- CTA specificity\n\n## Output contract\n\nReturn:\n\n1. creative map\n2. patterns associated with stronger/weaker outcomes\n3. mechanisms that may explain them\n4. confidence and confounders\n5. next creative hypotheses\n6. test matrix with success metric\n\nDo not recommend cloning superficial styling when the evidence points to the underlying proposition or proof structure instead.\n"
}