The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands (such as ls, cat, mkdir, and chmod) to manage the local .claude/skills directory, audit existing skills, and create new skill structures. It also demonstrates the use of dynamic context injection (!command) to retrieve real-time project information (e.g., git branch, gh pr diff) for agent prompts.
[SAFE]: The skill incorporates security-conscious patterns for secret management, explicitly instructing users to store credentials in a local .env file and use secure wrapper scripts to prevent API keys from being leaked into the conversation history. It also warns against the unintended execution of dynamic context and file references during documentation display.

create-agent-skills