codebase-map

Warn

Audited by Gen Agent Trust Hub on Jul 3, 2026

Risk Level: MEDIUMCOMMAND_EXECUTION
Full Analysis
  • [DYNAMIC_EXECUTION]: The skill executes a Python script at a hardcoded path (~/.claude/scripts/generate-codebase-map.py) that is not part of the skill package itself.
  • [INDIRECT_PROMPT_INJECTION]: The skill parses untrusted repository metadata including filenames and code signatures to generate a map that the agent subsequently reads as a codebase reference.
  • Ingestion points: The repository filesystem (SKILL.md).
  • Boundary markers: No explicit delimiters or warnings against embedded instructions are specified for the output file.
  • Capability inventory: Subprocess execution via python3 (SKILL.md).
  • Sanitization: No sanitization or validation of the ingested directory structure or code signatures is described.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jul 3, 2026, 05:44 AM
Security Audit — agent-trust-hub — codebase-map