using-git-worktrees
Pass
Audited by Gen Agent Trust Hub on Mar 12, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill performs shell operations using git commands for worktree management and invokes various language-specific package managers (npm, cargo, pip, poetry, go) to install dependencies and run tests within the new workspace.- [EXTERNAL_DOWNLOADS]: Installation steps for Node.js, Rust, Python, and Go projects trigger downloads from their respective official package registries (npmjs.org, crates.io, pypi.org, and proxy.golang.org) to fetch project dependencies.- [PROMPT_INJECTION]: The skill processes untrusted data from local repository files to determine its operational logic, which constitutes an indirect prompt injection risk.
- Ingestion points: The skill reads configuration preferences from
CLAUDE.mdand project metadata frompackage.json,Cargo.toml,requirements.txt,pyproject.toml, andgo.modin the current directory. - Boundary markers: No specific delimiters or safety instructions are used to separate ingested data from the agent's internal command logic.
- Capability inventory: The skill can execute arbitrary shell commands via package managers, modify the
.gitignorefile, commit changes to the repository, and perform network requests through build tools. - Sanitization: There is no evidence of sanitization or path validation for inputs derived from
CLAUDE.mdor project configuration files before they are used in path construction and command execution.
Audit Metadata