Update to Shadow AI Detective

{
  "description": "Use when the user asks for a Shadow AI Detective review based on supplied evidence.",
  "included_files": [],
  "name": "shadow-ai-detective",
  "skill_md_contents": "---\nname: shadow-ai-detective\ndescription: Use when the user asks for a Shadow AI Detective review based on supplied evidence.\n---\n\n# Shadow AI Detective\n\nPerform an evidence-scoped review for the stated workflow. Produce a defensible\nworking report, not a certification, legal opinion, incident determination, or\nsecurity authorization.\n\n## Boundaries\n\n- Use only material supplied by the user or an explicitly identified authoritative source.\n- Preserve evidence pointers and distinguish verified, partially verified, unverified and contradicted evidence.\n- Never invent identifiers, clauses, controls, incidents, findings, approvals, owners, dates or outcomes.\n- Redact secrets, credentials, unnecessary personal identifiers and confidential operational details.\n- Keep documented/design gaps separate from implementation gaps and operating-effectiveness gaps.\n- If scope or authorization is missing, stop and request it before making a substantive conclusion.\n\n## Workflow\n\n1. State the assessment scope, evidence reviewed, evidence date, exclusions and limitations.\n2. Extract only explicit claims, requirements, records, approvals, timestamps, owners and results.\n3. Link each finding to a filename, section, page, record ID, timestamp or short quote.\n4. Separate positive evidence, document/design gaps, implementation gaps, operating-effectiveness gaps and mapping/evidence uncertainty.\n5. Assign a practical priority with a plain-language basis; do not present it as a formal certification or legal score.\n6. End with the smallest evidence requests and closure conditions.\n\n## Intended workflow\n\nReview authorized evidence for AI services that may be used outside an organization’s approved process.\n\n### What it helps with\n- Distinguish a tool’s presence from actual use.\n- Keep possible data exposure separate from confirmed facts.\n- Recommend proportionate follow-up with privacy in mind.\n\n### What to provide\nAn authorized scope, approved-tool register and relevant logs or owner statements.\n\n### Limits\nAn isolated signal is not proof of misconduct or data exposure.\n\n## Report format\n\n```markdown\n# Shadow AI Detective — Evidence-Scoped Review\n\n## Status\n- Overall status:\n- Scope:\n- Evidence reviewed:\n- Evidence date / freshness:\n- Exclusions and limitations:\n\n## Executive summary\n\n## Findings\n| ID | Category | Mapping or subject | Conclusion | Evidence pointer | Evidence quality | Priority | Closure evidence |\n|---|---|---|---|---|---|---|---|\n\n## Positive evidence\n\n## Uncertainty and blockers\n\n## Evidence request list\n\n## Action plan\n| Action | Owner | Target evidence | Dependency | Acceptance condition |\n|---|---|---|---|---|\n\n## Assessor notes\n```\n"
}