The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on various shell commands to perform git and GitHub CLI operations.
Executed commands include git worktree, git branch, gh issue, and gh pr across the create.md, remove.md, and prune.md files.
[CREDENTIALS_UNSAFE]: The skill identifies and accesses sensitive environment files during the worktree creation process.
File create.md contains logic to search for .env, .env.local, and .envrc files using find operations.
The skill includes a procedure to prompt the user for confirmation before copying these files to the new worktree directory to ensure the user is aware of secret handling.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection via untrusted data retrieved from external sources like GitHub issues and pull requests.
Ingestion points: In create.md, the skill fetches issue and PR titles using the GitHub CLI (gh).
Boundary markers: There are no explicit boundary markers or instructions to the agent to ignore embedded commands within the fetched issue/PR data.
Capability inventory: The skill performs file system operations and shell command execution (git worktree add, git checkout, etc.) which could be targeted by malicious data.
Sanitization: The skill implements sanitization in create.md (lines 42-44) using sed and tr to filter the title into a safe alphanumeric slug before using it in directory paths and branch names.

worktree