skills/desek/governance/reconcile-cr/Gen Agent Trust Hub

reconcile-cr

Pass

Audited by Gen Agent Trust Hub on Oct 30, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill identifies and processes information from iteration ledgers and git logs to update project documentation, which presents a potential surface for indirect prompt injection from untrusted ledger content.
  • Ingestion points: The skill reads external data from ledger files in docs/cr/iterations/ and commit data via git log (SKILL.md, Step 1).
  • Boundary markers: A mandatory reconciliation plan must be approved by the user in Step 3 before any file modifications occur, serving as a critical safety checkpoint.
  • Capability inventory: The skill is restricted to writing documentation files in the docs/ folder and executing git commit for those specific paths (SKILL.md, Steps 4-8).
  • Sanitization: The combination of explicit safety rules (prohibiting code changes) and the human review process effectively mitigates the risk of processing malicious instructions embedded in the source data.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 30, 2026, 06:49 PM