using-git-worktrees
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill performs standard shell operations for repository management and development workflows. This includes git commands for worktree management (
git worktree add,git rev-parse), filesystem checks (ls,cd), and the invocation of common package managers and test runners. - [INDIRECT_PROMPT_INJECTION]: The skill identifies the project type by reading filenames within the workspace to determine which setup and test commands to run. This represents a potential surface where malformed repository structures could influence agent behavior, though the actions taken are limited to standard development tools.
- Ingestion points: The skill checks for the existence of
package.json,Cargo.toml,requirements.txt,pyproject.toml, andgo.modin the current directory (SKILL.md). - Boundary markers: None explicitly defined; the agent relies on the existence of these files to choose a branch of logic.
- Capability inventory: The skill utilizes subprocess calls for
npm,cargo,pip,poetry, andgocommands across Step 2 and Step 3. - Sanitization: No explicit sanitization of file contents is performed, but the commands executed are strictly limited to well-known package manager installation and test flags.
Audit Metadata