The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the Bash, Glob, and Grep tools to perform an inventory of agent files within the project directory (.claude/agents/), the user's home directory (~/.claude/agents/), and the plugin scope (brewcode/agents/). This behavior is integral to the skill's function of managing installed agents and identifying configuration health issues.
[REMOTE_CODE_EXECUTION]: The skill orchestrates complex tasks by spawning specialized subagents using the Agent tool (e.g., brewcode:agent-creator). These subagents are invoked with clearly defined goals and scopes to ensure isolation and bounded execution. All delegated subagents belong to the same vendor ecosystem.
[DATA_EXPOSURE]: The skill reads agent definitions and system prompts to perform quality reviews and synchronization. It handles metadata like trigger keywords and model fields. This data access is restricted to the specific directories intended for agent configuration.

agents