refactor
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill functions by ingesting and refactoring untrusted code fragments, which creates a potential surface for indirect prompt injection. 1. Ingestion points: Code snippets provided in the $ARGUMENTS section or read from repository files. 2. Boundary markers: There are no explicit delimiters or instructions provided to separate the input code from the agent's internal logic. 3. Capability inventory: The skill is intended to use the agent's abilities to read, analyze, and output or modify code files. 4. Sanitization: No sanitization or validation mechanisms are mentioned to check the input code for embedded malicious instructions.
Audit Metadata