Update to Marketing Swarm

{
  "description": "Use when the user has prior campaign results, experiments, creative history, or benchmarks and wants to find comparable past evidence, reusable patterns, or lessons relevant to a current decision.",
  "included_files": [
    {
      "relative_path": "agents/openai.yaml",
      "size_in_bytes": 298
    }
  ],
  "name": "marketing-memory",
  "skill_md_contents": "---\nname: marketing-memory\ndescription: Use when the user has prior campaign results, experiments, creative history, or benchmarks and wants to find comparable past evidence, reusable patterns, or lessons relevant to a current decision.\n---\n\n# Marketing Memory\n\nUse the user's own historical evidence as context without turning anecdotes into universal rules.\n\n## Retrieve by decision relevance\n\nMatch past cases using dimensions that matter to the current question, such as:\n\n- platform and campaign objective\n- market / geography\n- product, price, margin, or offer\n- audience intent and size\n- funnel stage\n- budget scale\n- seasonality / promotion state\n- creative format and message angle\n- attribution setup\n- measurement quality\n\nDo not rank cases by superficial similarity when economically important conditions differ.\n\n## Convert history into evidence\n\nFor each relevant case capture:\n\n- context\n- action or treatment\n- observed result\n- measurement quality\n- what changed at the same time\n- lesson\n- transferability to the current case\n\nDistinguish:\n\n- repeated pattern\n- single precedent\n- failed test\n- unresolved case\n\n## Benchmark rule\n\nUse external benchmarks only when the user explicitly provides them or the host can retrieve current credible sources. Never invent industry averages. User-specific historical baselines are often more useful than generic benchmarks.\n\n## Output contract\n\nReturn the most relevant prior cases, why they match, why they may not transfer, the lesson to reuse, and the current hypothesis they support or weaken.\n\nWhen no sufficiently comparable history exists, say so and route to `scenario-simulation` or an experiment rather than forcing a memory match.\n"
}