security-review
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's core functionality involves reading and analyzing untrusted project files (source code, configuration, and logs), which creates a surface for indirect prompt injection attacks. Malicious instructions hidden within the scanned codebase could attempt to influence the agent's reasoning or report accuracy.
- Ingestion points: The skill reads arbitrary codebase files during the resolution and deep scan steps (SKILL.md, Steps 1, 3, 4, and 5).
- Boundary markers: The skill incorporates a 'Self-Verification Pass' and a 'Zero-Trust Verification' protocol to validate findings, but it lacks strict prompt delimiters or instructions to ignore embedded directives in the scanned content.
- Capability inventory: The agent analyzes code and generates textual security reports and proposed patches. It does not execute the code it audits or make external network requests.
- Sanitization: The skill does not perform automated sanitization of the untrusted code content before processing it for analysis.
Audit Metadata