code-quality
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes and modifies repository code, which represents an indirect prompt injection surface. The instructions mitigate this risk by requiring the agent to maintain security controls and validate data boundaries.
- Ingestion points: The skill operates on repository files identified via user-named scope, changed files, or branch diffs.
- Boundary markers: It explicitly instructs the agent to treat data at untrusted boundaries as unknown and to narrow or parse it there.
- Capability inventory: The skill is designed to review and implement code changes, involving file modification capabilities.
- Sanitization: Instructions mandate the preservation of trust-boundary validation, authorization, data integrity, and error handling.
Audit Metadata