metal

Pass

Audited by Gen Agent Trust Hub on Mar 18, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted data from arbitrary codebases, creating an indirect prompt injection surface.
  • Ingestion points: Reads repository source code, documentation, and configuration files.
  • Boundary markers: No explicit delimiters or instructions are used to separate analyzed data from the agent's instructions.
  • Capability inventory: Utilizes Read, Grep, Glob, and Bash tools to examine repository contents.
  • Sanitization: No evidence of validation or sanitization of the content extracted from external files is present.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 18, 2026, 07:14 AM