promote
Pass
Audited by Gen Agent Trust Hub on Aug 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill requires the agent to ingest and interpret repository metadata, including
READMEfiles,Makefiles, and CI workflow definitions, to discover project-specific promotion flows. While this exposes a surface for instructions embedded in source code to influence agent logic, the skill provides a rigid framework of verification steps (drift checks, tree-SHA comparisons) and manual gates to maintain control.\n - Ingestion points: Files like
README,CLAUDE.md,Makefile, and.github/workflows/are accessed in Step 0.\n - Boundary markers: The skill does not define specific markers for untrusted content but mandates a 'write down the discovered flow' step for human verification.\n
- Capability inventory: Employs standard developer tools including
git(diff, rev-parse, fetch),make, and CI dispatch commands.\n - Sanitization: None; relies on procedural verification and human approval gates.\n- [SAFE]: The skill follows security best practices by discouraging ad-hoc command execution, requiring human-in-the-loop approvals for production changes, and warning against common pitfalls like secret exposure and unauthorized privilege escalation.
Audit Metadata