security-audit
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEPROMPT_INJECTIONCREDENTIALS_UNSAFE
Full Analysis
- [PROMPT_INJECTION] (LOW): Potential for Indirect Prompt Injection. The skill is designed to ingest and analyze untrusted code snippets and directory contents. This creates a vulnerability surface where instructions embedded in analyzed data could potentially influence agent behavior. * Ingestion points: Analyzes user-provided code blocks and directory paths (SKILL.md). * Boundary markers: Missing explicit delimiters or instructions to ignore commands within the data being analyzed. * Capability inventory: No dangerous operations (network or file-write) are defined in the skill instructions themselves. * Sanitization: No input sanitization or filtering logic is provided.
- [CREDENTIALS_UNSAFE] (SAFE): The skill includes a synthetic placeholder API key (sk_live_abc123xyz) used for demonstration and detection purposes within its documentation. This is not a functional or sensitive credential.
Audit Metadata