The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local shell script scripts/claude-wt.sh for all its primary functions. The content and logic of this script are not provided, preventing verification of its safety, handling of user-supplied arguments, or potential side effects on the host system.
[CREDENTIALS_UNSAFE]: The procedure automatically copies .env.local files to directories at ../worktrees/. As .env files typically contain secrets, API keys, and other credentials, this programmatic movement of sensitive data to unmanaged locations outside the project root increases the risk of credential exposure.
[EXTERNAL_DOWNLOADS]: The skill automatically runs npm install in newly created worktrees. This triggers the download and execution of third-party code from external package registries, which poses a supply chain risk if the target branch or its dependencies are compromised.
[COMMAND_EXECUTION]: The skill processes user-supplied $ARGUMENTS into shell commands. While the new command mentions sanitization to a slug, other subcommands such as remove <name> do not explicitly state sanitization steps, potentially exposing the system to command injection if the underlying script handles the input unsafely.

worktree