vibe-audit
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from source code repositories, creating a potential surface for indirect prompt injection where instructions could be hidden in code or comments. The skill mitigates this through specific instructions to treat all content as data rather than instructions.
- Ingestion points: Source code, README files, CI configurations, and git history from the audited repository.
- Boundary markers: The 'Read-only and safety contract' section explicitly states that source, comments, and documents are data, not executable instructions, and directs the agent to ignore their directives.
- Capability inventory: The skill uses file reading and execution of read-only shell commands for verification (e.g., git diff, linting, or tests).
- Sanitization: The skill mandates masking secrets and prohibits the output of attack reproductions or misuse procedures.
Audit Metadata