linking
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill analyzes source code, which acts as an ingestion point for untrusted data. While it presents an attack surface for indirect prompt injection, its capabilities are restricted to local file analysis.
- Ingestion points: Reads local source code, configuration files, and database schemas resolved via the
--scopeflag in SKILL.md. - Boundary markers: No explicit delimiters or instructions are provided to the agent to treat the analyzed code as data distinct from instructions.
- Capability inventory: Limited to file reading and pattern matching (grep) within the local workspace.
- Sanitization: No sanitization of the analyzed content is described.
- [SAFE]: The skill is a diagnostic tool for identifying privacy vulnerabilities. It does not perform network operations to untrusted domains, execute remote code, or request elevated system privileges. All analysis is performed on local files within the agent's scoped workspace.
Audit Metadata