reorganize

Pass

Audited by Gen Agent Trust Hub on Aug 15, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests data from local project files that may contain untrusted content to generate its reorganization proposals.
  • Ingestion points: Processes project-specific metadata and documentation (e.g., AGENTS.md, CONTRIBUTING, package.json) and source code hotspots as defined in the SKILL.md 'First Steps' section.
  • Boundary markers: The instructions do not mandate the use of delimiters or instructions to ignore embedded commands when the agent reads external file content.
  • Capability inventory: The agent has permissions to glob the filesystem, read file contents, and perform file system writes (moves and extractions) in the 'Apply' workflow stage.
  • Sanitization: No explicit sanitization, filtering, or validation is performed on the content extracted from the project files.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 15, 2026, 04:37 PM
Security Audit — agent-trust-hub — reorganize