loop-on-ci
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes potentially untrusted data from external CI environments which could contain malicious instructions designed to influence the agent's behavior during the fix process.
- Ingestion points: The skill reads PR check metadata via
gh pr checksand detailed failure logs usinggh run view --log-failed. - Boundary markers: No explicit delimiters or instructions are provided to the agent to treat external log content as non-instructional data.
- Capability inventory: The skill instructions empower the agent to diagnose failures and apply fixes, which implies the ability to write to the filesystem and perform git operations (push/commit) to resolve CI errors.
- Sanitization: There is no evidence of sanitization or filtering of the CI logs before they are provided to the agent for analysis.
Audit Metadata