close-github-worktree-done

Pass

Audited by Gen Agent Trust Hub on Aug 14, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes local git commands such as git worktree list, git branch, and git fetch to manage the local development environment. It also invokes local PowerShell and Bash scripts located in the skill directory (.cursor/skills/github-worktree/scripts/) to perform worktree removal tasks.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests data from the local repository state and user input to determine target directories and parse GitHub issue numbers.
  • Ingestion points: Branch names are retrieved from git worktree list or provided directly by the user.
  • Boundary markers: The skill uses double quotes to wrap shell variables (e.g., "<TARGET_ROOT>" and "<目标分支名>") when executing commands, which helps prevent basic command injection.
  • Capability inventory: File system modifications (removing worktrees and branches) and network operations (git fetch and GitHub API calls via the user-github MCP server).
  • Sanitization: The logic includes specific parsing for numeric issue IDs and mandates explicit user confirmation before deleting worktrees with uncommitted changes or branches that haven't been merged into master.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 14, 2026, 06:41 AM
Security Audit — agent-trust-hub — close-github-worktree-done