The Agent Skills Directory

[SAFE]: The skill is primarily instructional documentation and prompt templates for the Claude Code routines feature. It includes extensive safety checklists and rules for the agent to follow to ensure generated prompts are secure and functional.
[DATA_EXFILTRATION]: The skill implements strong data protection practices by explicitly forbidding the embedding of real API tokens in generated output and providing templates that use environment variable placeholders (e.g., $ROUTINE_FIRE_TOKEN).
[COMMAND_EXECUTION]: Provides utility script templates for checking the local environment's compatibility with the claude CLI. These scripts are diagnostic and restricted to verifying the presence and basic functionality of the official tool.
[PROMPT_INJECTION]: Features a dedicated 'AI Self-Check' section that instructs the agent to verify that generated prompts are self-contained, include success criteria, and maintain safety boundaries, which acts as a defense against unintended behavior.

routine-writer