git-worktrees-usage

Warn

Audited by Gen Agent Trust Hub on Feb 21, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • COMMAND_EXECUTION (MEDIUM): The skill automatically executes package management, build, and test commands based on detected project files.
  • Evidence: The skill scripts execute npm install, cargo build, pip install, poetry install, go mod download, npm test, cargo test, pytest, and go test automatically upon creating a worktree.
  • Risk: Malicious repositories can include harmful code in install scripts (e.g., postinstall in package.json) or test suites. Because the skill triggers these automatically, it bypasses manual review of the repository's safety.
  • EXTERNAL_DOWNLOADS (LOW): Execution of package managers results in downloading third-party code from external registries (NPM, PyPI, etc.).
  • PROMPT_INJECTION (LOW): The skill ingests untrusted data from AGENTS.md and project configuration files to determine its behavior.
  • Ingestion points: AGENTS.md (via grep), package.json, Cargo.toml, requirements.txt, pyproject.toml, go.mod.
  • Boundary markers: None. The skill assumes the contents of these files are safe and intended for configuration.
  • Capability inventory: Full subprocess execution for package managers and testing frameworks.
  • Sanitization: None. The skill executes commands directly based on the presence of these files.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 21, 2026, 04:56 PM