macs

Use when supervising a worker agent in another terminal. Loads MACS controller operating rules and decision priorities.

Use for BMAD BMB (builder module) tasks: creating/editing BMAD agents, workflows, or modules. Trigger on BMAD builder, agent builder, workflow builder, or module creation requests.

Use for BMAD BMGD (game development) workflows, game briefs/GDDs, and game-specific planning. Trigger on BMGD, game design, GDD, or game architecture requests.

Use for BMAD BMM (BMad Method) software-dev workflows: PRD, architecture, epics/stories, sprint planning, quick-spec/quick-dev, and lifecycle tracking. Trigger on 'BMM', 'BMad Method', or when users want structured planning/implementation via BMAD.

Use for BMAD CIS (Creative Intelligence Suite) tasks like brainstorming, innovation strategy, problem solving, or storytelling. Trigger on CIS, creative suite, brainstorming coach, or storytelling workflows.

Use when a user says 'use BMAD', 'follow BMAD', or wants the BMAD Method run in a repo. Routes BMAD work to the worker terminal, verifies install, and selects the right BMAD module/track.

Use for BMAD TEA (Test Architect) workflows: test-design, atdd, automate, test-review, nfr-assess, trace, and test framework/CI setup. Trigger on TEA, Test Architect, or BMAD testing/quality gates.

Keep the controller looping and progressing work without interruption.

Aider runtime/session control: use Aider CLI slash commands and flags to manage files, models, and sessions in a worker terminal. Trigger when the controller needs to operate the Aider CLI like a human.

Backend Developer agent for Codex multi-agent workflows. Use when implementing server endpoints from REQUIREMENTS.md and AGENT_TASKS.md; outputs /backend/server.js and /backend/package.json.

Claude Code runtime/session control: use Claude Code slash commands and CLI flags to manage sessions, permissions, model selection, and restarts in a worker terminal. Trigger when the controller needs to operate the Claude Code CLI like a human.

Codex runtime/session control: use Codex CLI slash commands and CLI flags to manage sessions, approvals, models, resumes/forks, and safe restarts in a worker terminal. Trigger when the controller needs to operate the Codex CLI like a human.

Designer agent for Codex multi-agent workflows. Use when creating implementation-ready UI specs and optional wireframes based on REQUIREMENTS.md and AGENT_TASKS.md; outputs /design/design_spec.md (and /design/wireframe.md).

Frontend Developer agent for Codex multi-agent workflows. Use when implementing the UI from /design/design_spec.md and AGENT_TASKS.md; outputs /frontend/index.html, /frontend/styles.css, /frontend/main.js.

Gemini CLI runtime/session control: use Gemini CLI slash commands and CLI flags to manage sessions, models, and restarts in a worker terminal. Trigger when the controller needs to operate the Gemini CLI like a human.

llama.cpp runtime/session control: use llama-cli and llama-server commands/flags to run local GGUF models and serve an API in a worker terminal. Trigger when the controller needs to operate llama.cpp like a human.

LM Studio runtime/session control: use lms CLI commands to manage models and servers in a worker terminal. Trigger when the controller needs to operate LM Studio CLI like a human.

Ollama runtime/session control: use ollama CLI commands to run models, manage downloads, and stop running models in a worker terminal. Trigger when the controller needs to operate Ollama like a human.

Open Interpreter runtime/session control: use Open Interpreter CLI % commands and shell flags to manage sessions and context in a worker terminal. Trigger when the controller needs to operate Open Interpreter like a human.

Project Manager agent for Codex multi-agent workflows. Use when coordinating Designer, Frontend, Backend, and Tester roles; when you must produce REQUIREMENTS.md, TEST.md, and AGENT_TASKS.md; or when you need to manage agent handoffs and acceptance criteria.

Tester agent for Codex multi-agent workflows. Use when validating outputs against TEST.md and AGENT_TASKS.md; outputs /tests/TEST_PLAN.md (and /tests/test.sh if needed).