The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several bundled shell scripts (detect-mode.sh, verify-team.sh, toggle-team.sh, trace-ops.sh) to handle environment detection, integrity verification, and state management within the project's .claude/ directory.
[REMOTE_CODE_EXECUTION]: Spawns specialized sub-agents (such as Explore and brewcode:agent-creator) using the Task tool to perform project auditing and instruction generation. This is a standard delegated execution pattern for complex autonomous workflows.
[DYNAMIC_EXECUTION]: The skill dynamically generates and updates AI agent instruction files (.claude/agents/*.md) by assembling content from templates, user prompts, and project analysis results. This behavior is the intended primary purpose of the skill and is managed via structured templates.
[INDIRECT_PROMPT_INJECTION]: The skill ingests project data (tech stack, architectural layers, and documentation) to define agent domains. It employs explicit boundary instructions and "Return Contracts" to manage the reliability of generated agents and minimize risks from processing external project content.

teams-setup