fix-ci
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources which could contain malicious instructions designed to influence the agent's logic during the diagnostic phase.
- Ingestion points: GitHub Actions job logs (via
gh run view --job <job-id> --log-failed) and pull request diffs. - Boundary markers: The instructions do not specify delimiters or markers to separate log content from the agent's internal processing logic.
- Capability inventory: The skill possesses significant capabilities including local file system modification (
git worktree), code commits, and pushing changes to remote repositories (git push). - Sanitization: No explicit sanitization or filtering of log content is performed before the agent interprets the cause of failure.
- [COMMAND_EXECUTION]: The skill utilizes the GitHub CLI (
gh) and Git to perform its primary functions. It also executes local build and test commands (e.g.,go test,go vet) to verify fixes. - Evidence: Use of
gh run,gh pr,git worktree, and repository-specific test runners.
Audit Metadata