Update to 8gnc — Brand Growth Diagnostic

{
  "name": "diagnose-brand-growth",
  "description": "Diagnose ambiguous brand and growth problems and route them to the smallest useful 8gnc method sequence. Use when the user says the brand feels stuck, generic, invisible, inconsistent, hard to explain, weak at conversion, weak in outreach, unclear on product direction, or asks which 8gnc skill to use. Also trigger for broad requests such as \"fix my brand,\" \"what is actually wrong,\" \"where should we start,\" \"audit this business,\" or \"give me the full 8gnc treatment.\" Do not trigger when the user explicitly names a specialist skill and already has the inputs it requires.",
  "included_files": [
    {
      "relative_path": "agents/openai.yaml",
      "size_in_bytes": 268
    },
    {
      "relative_path": "references/capability-map.md",
      "size_in_bytes": 3687
    }
  ],
  "skill_md_contents": "---\nname: diagnose-brand-growth\ndescription: Diagnose ambiguous brand and growth problems and route them to the smallest useful 8gnc method sequence. Use when the user says the brand feels stuck, generic, invisible, inconsistent, hard to explain, weak at conversion, weak in outreach, unclear on product direction, or asks which 8gnc skill to use. Also trigger for broad requests such as \"fix my brand,\" \"what is actually wrong,\" \"where should we start,\" \"audit this business,\" or \"give me the full 8gnc treatment.\" Do not trigger when the user explicitly names a specialist skill and already has the inputs it requires.\n---\n\n# Diagnose Brand and Growth\n\nStart with diagnosis. Do not expose a 36-skill menu or run the entire stack by default.\n\nRead [capability-map.md](references/capability-map.md) before routing. Load only the selected specialist skill instructions after the route is chosen.\n\n## 1. Establish the evidence boundary\n\nCollect or inspect the smallest useful evidence set:\n\n- the decision the user needs to make;\n- the audience, offer, market, and current surface;\n- available first-party evidence such as interviews, analytics, loss reasons, briefs, drafts, or pages;\n- constraints, timing, and what has already been tried.\n\nLabel every input as observed, user-reported, inferred, or unknown. Never invent client intelligence, market evidence, metrics, approvals, or access.\n\nThis release has no authenticated client-data connection. Use only evidence the user provides, files available in the current workspace, and lawful public research. Treat an approval as a decision about the named artifact, not authorization to publish, deploy, send, or execute.\n\n## 2. Name the primary constraint\n\nChoose one primary lane:\n\n- brand meaning or positioning;\n- product direction or sequencing;\n- content or voice;\n- search or AI visibility;\n- conversion or pricing architecture;\n- sales motion or practice.\n\nUse a secondary lane only when evidence shows it is upstream or downstream of the primary constraint. Distinguish symptoms from causes. Low traffic is not automatically an SEO problem; weak conversion is not automatically a copy problem; generic content is not automatically a voice problem.\n\n## 3. Route the work\n\nSelect the minimum sequence from the capability map. Prefer:\n\n- one diagnostic or research method;\n- one synthesis or stress-test method;\n- one deliverable method only when the inputs are strong enough.\n\nDo not run compilers before their required layer artifacts exist. Do not call simulated audience reactions real research. Do not call speculative recommendations validated strategy.\n\nIf the user explicitly requests a named specialist skill, invoke that skill directly unless a missing prerequisite would make the output misleading.\n\n## 4. Return a decision-ready diagnosis\n\nReturn:\n\n1. **Primary constraint** — one sentence.\n2. **Evidence** — what supports it, with source labels.\n3. **Confidence and gaps** — high, medium, or low, plus what would change the call.\n4. **Route** — the selected skills in order and why each is necessary.\n5. **First move** — the first concrete input or action, not a generic checklist.\n6. **Decision gate** — what a human must decide before implementation.\n\nEnd substantial outputs with these five lines:\n\n- Evidence supports:\n- Evidence does not establish:\n- Human judgment required:\n- Implementation prerequisites:\n- The model cannot know:\n\nDo not end with an agency pitch. The method must be useful on its own.\n"
}