Update to Marketing Swarm

{
  "name": "marketing-swarm-router",
  "description": "Use for broad or multi-part paid-media questions that need routing across campaign diagnosis, budget allocation, creative analysis, fatigue planning, simulation, attribution, historical evidence, and quality review.",
  "included_files": [
    {
      "relative_path": "agents/openai.yaml",
      "size_in_bytes": 281
    }
  ],
  "skill_md_contents": "---\nname: marketing-swarm-router\ndescription: Use for broad or multi-part paid-media questions that need routing across campaign diagnosis, budget allocation, creative analysis, fatigue planning, simulation, attribution, historical evidence, and quality review.\n---\n\n# Marketing Swarm Router\n\nRoute a marketing question to the smallest useful set of specialist Skills. Do not answer every request with every Skill.\n\n## First classify the job\n\nIdentify one or more of these jobs:\n\n- **diagnose**: performance drop, pacing, spend anomaly, account health, delivery instability -> `campaign-diagnostics`\n- **allocate**: budget split, platform mix, marginal spend, media opportunity -> `budget-media-allocation`\n- **decode creative**: understand why ads differ, identify hook/promise/proof/CTA patterns -> `creative-genome-analysis`\n- **refresh creative**: fatigue, decay, rotation, next variants, test matrix -> `creative-fatigue-mutation`\n- **forecast**: what-if, sensitivity, expected ranges, budget scenarios -> `scenario-simulation`\n- **test causality**: attribution, incrementality, lift, confounders, counterfactual claims -> `causal-attribution`\n- **compare history**: past campaigns, benchmarks from supplied history, reusable lessons -> `marketing-memory`\n- **review decision**: check evidence quality, contradictions, risk, overclaiming -> `decision-quality-gate`\n\n## Evidence inventory\n\nBefore routing, identify what is actually available:\n\n- business objective and primary KPI\n- platform/account/campaign/ad set/creative grain\n- reporting period and comparison period\n- spend, impressions, clicks, conversions, revenue or value\n- attribution window/model where relevant\n- creative identifiers and launch dates where relevant\n- prior tests, holdouts, experiments, or historical campaigns\n- known constraints: budget floors, inventory, geography, policy, learning phase, margin, capacity\n\nDo not block on every missing field. Continue with the evidence available and state which missing fields materially lower confidence.\n\n## Routing principles\n\n1. Use the narrowest specialist that can solve the job.\n2. Parallelize independent analyses when the host supports it, but reconcile them before answering.\n3. Do not treat platform-reported attribution as causal proof.\n4. Do not treat a forecast as observed evidence.\n5. Do not recommend a budget shift without checking whether the apparent winner is affected by volume, learning, attribution, inventory, or creative fatigue.\n6. For high-impact decisions, invoke `decision-quality-gate` before finalizing.\n\n## Standard answer frame\n\nFor multi-Skill work, return:\n\n- **What changed**: the observed pattern\n- **Most likely explanations**: ranked, with evidence for/against\n- **What is uncertain**: missing evidence or confounders\n- **What to do next**: prioritized actions/tests\n- **Decision thresholds**: what result would make you continue, stop, scale, or reverse\n\nIf a compatible host app can make campaign changes, analysis and execution remain separate. Never turn a recommendation into a live mutation without explicit user authorization for the material change.\n"
}