The Agent Skills Directory

[SAFE]: The skill functions as a static mapping for intent-based task delegation, guiding the agent to the correct sub-skills (e.g., design, debugging, or research) without executing dangerous operations.
[COMMAND_EXECUTION]: The skill provides a bash command intended for the agent to verify that its internal routing table is synchronized with the installed skill-lock.json. This command uses standard, non-malicious tools (rg, jq, diff) on local files to ensure system integrity.
[SAFE]: References to internal configuration files like ~/.claude/settings.json and ~/.agents/.skill-lock.json are used solely for environment verification and plugin status checks, which is standard behavior for this platform's skill management.

using-skills