model-audit
Pass
Audited by Gen Agent Trust Hub on Mar 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [SAFE]: No malicious patterns or security risks were identified. The skill's behavior is consistent with its stated purpose of auditing code for model routing and pricing.
- [COMMAND_EXECUTION]: The skill uses recursive grep commands to locate model configurations and API calls within the codebase. These commands are restricted to standard source file extensions (.py, .ts, .js) and do not involve elevated privileges.
- [PROMPT_INJECTION]: The skill processes untrusted local codebase data, which represents an indirect prompt injection surface. Evidence: 1. Ingestion points: Local source files (.py, .ts, .js) identified via grep patterns. 2. Boundary markers: Absent; there are no instructions to the agent to ignore potential instructions within the analyzed code. 3. Capability inventory: File reading and shell search command execution. 4. Sanitization: Absent; the skill does not filter or sanitize the content of the files before processing.
Audit Metadata