using-git-worktrees
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses git commands and standard shell utilities like ls and grep to manage isolated workspaces and detect configuration preferences.
- [COMMAND_EXECUTION]: It automates project setup by executing package manager commands including npm install, cargo build, pip install, and go mod download based on detected project files.
- [INDIRECT_PROMPT_INJECTION]: The skill identifies a potential attack surface by ingesting data from project-local files like AGENTS.md and CLAUDE.md.
- Ingestion points: Project configuration and instruction files including AGENTS.md, CLAUDE.md, and standard package manifests.
- Boundary markers: Includes a comprehensive untrusted-content-contract defining how the agent should treat external data and ignore embedded instructions.
- Capability inventory: Executes git, package managers (npm, pip, poetry, cargo, go), and testing frameworks.
- Sanitization: Relies on instructional delimiters and explicit safety policies defined in the skill body to prevent the execution of instructions found within external project data.
Audit Metadata