investigate
Pass
Audited by Gen Agent Trust Hub on Mar 27, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is designed for read-only exploration and documentation of existing features. It includes a robust 'Evidence-Based Reasoning Protocol' and a 'Golden Rule' that requires the agent to verify all claims with specific
file:lineproof before reaching conclusions, which significantly reduces the risk of incorrect or harmful recommendations. - [PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection because it is designed to ingest and analyze untrusted codebase content.
- Ingestion points: Tools used to read and grep codebase files (e.g., Read, Grep, Glob).
- Boundary markers: The instructions do not define explicit delimiters for untrusted file content, but they do mandate the use of shared protocols that emphasize skepticism and validation.
- Capability inventory: The skill allows reading, searching, and writing analysis artifacts (via
TodoWrite), but explicitly forbids making code changes. - Sanitization: No explicit sanitization of file content is described; however, the requirement for confidence levels and evidence-based verification acts as a procedural safeguard against malicious instructions hidden in code comments.
Audit Metadata