atelier-review-me

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted input from code diffs and PRs but includes robust instructions to treat this content as data rather than instructions.
  • Ingestion points: reads git diffs and PR metadata (SKILL.md).
  • Boundary markers: 'Untrusted input' section provides clear prompt boundaries and instructions to ignore embedded commands (SKILL.md).
  • Capability inventory: performs git operations and script execution (SKILL.md).
  • Sanitization: instructs the agent to report injections instead of executing them (SKILL.md).
  • [COMMAND_EXECUTION]: Uses standard development tools like git diff to identify changes in the codebase for analysis.
  • [DYNAMIC_EXECUTION]: Executes a local project script scripts/trigger-eval/run.sh to validate configuration changes when skill definitions are modified.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 04:06 PM
Security Audit — agent-trust-hub — atelier-review-me