Update to Adszy

{
  "name": "account-health-triage",
  "description": "Diagnose the current health of an authorised Google Ads account through Adszy MCP. Use for account check-ups, \"what needs attention\", budget-pacing, disapproval, anomaly, audit, or sudden-performance questions.",
  "included_files": [
    {
      "relative_path": "agents/openai.yaml",
      "size_in_bytes": 224
    }
  ],
  "skill_md_contents": "---\nname: account-health-triage\ndescription: Diagnose the current health of an authorised Google Ads account through Adszy MCP. Use for account check-ups, \"what needs attention\", budget-pacing, disapproval, anomaly, audit, or sudden-performance questions.\n---\n\n# Account Health Triage\n\nUse Adszy's read-only tools to separate verified issues from healthy or unobserved areas.\n\n## Workflow\n\n1. If the account is ambiguous, call `list_ad_accounts` and ask the user to choose. Never guess an account.\n2. Establish the requested window. Use `get_account_summary` for headline performance, then call only the relevant diagnostic tools:\n   - `get_budget_pacing` for spend delivery.\n   - `get_anomalies` for detected changes.\n   - `get_disapprovals` for serving restrictions.\n   - `get_account_audit` for the latest cached structural audit.\n3. Treat a cached audit window as authoritative. If `windowMismatch` is true, state its actual window and use `get_account_summary` for the user's requested period.\n4. Distinguish:\n   - verified issue;\n   - healthy observed surface;\n   - no data or unvalidated surface.\n5. Return a prioritised brief: issue, evidence, likely impact, and next safe action.\n\n## Guardrails\n\n- Never claim Adszy changed the account; every MCP tool is read-only.\n- Do not infer causation from correlation. Label plausible explanations as hypotheses.\n- State dates, currency, and material sample limitations.\n- Do not describe an empty result as healthy.\n- Keep account identifiers out of the prose unless the user needs them to disambiguate.\n"
}