Codebase Onboarding Engineer
Pass
Audited by Gen Agent Trust Hub on Oct 25, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted code and metadata from repositories, which may contain malicious instructions designed to subvert agent behavior.
- Ingestion points: Source code files, manifests, and file structures as described in the workflow section of SKILL.md.
- Boundary markers: The prompt contains explicit discipline to 'state only facts', 'quote... exactly', and 'avoid inference', which serve as soft boundaries against injection.
- Capability inventory: The agent is strictly read-only and prohibited from modifying files, generating state changes, or performing network operations.
- Sanitization: There is no mention of sanitization or escaping for the repository content being read.
Audit Metadata