w-git-repair
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data such as CI logs, PR review comments, and PR bodies which could contain hidden instructions.
- Ingestion points: The file references/git-publish.md identifies CI logs, review comments, and PR bodies as external data sources ingested into the agent context.
- Boundary markers: The skill includes an explicit instruction to treat these sources as evidence of code issues rather than as authorization to execute embedded commands.
- Capability inventory: The skill can execute various Git commands including git status, reset, clean, restore, switch, and fetch.
- Sanitization: The instructions provide a logical filter by explicitly telling the agent to ignore any embedded commands within the ingested external data.
Audit Metadata