init-agents-md

Pass

Audited by Gen Agent Trust Hub on Apr 12, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill's core functionality is limited to reading local codebase metadata and project configuration files to generate a documentation draft. It does not perform network requests or execute external scripts.\n- [PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection by ingesting data from arbitrary project files to generate instructions. However, this is mitigated by a mandatory human-in-the-loop review process.\n
  • Ingestion points: Reads various project files including package.json, pyproject.toml, README.md, and existing AGENTS.md (SKILL.md, Step 1).\n
  • Boundary markers: The instructions do not define specific markers to delimit ingested file content from the agent's task instructions.\n
  • Capability inventory: The skill utilizes codebase exploration and file-writing capabilities to generate project root documentation (SKILL.md, Steps 1 & 5).\n
  • Sanitization: No content validation is specified for ingested data, but the draft iteration phase (Step 4) requires explicit user approval before final file creation.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 12, 2026, 11:08 PM