| name | collaboration |
| description | Multi-AI collaboration for tasks requiring multiple perspectives, discussion, or negotiation. Use when: user explicitly requests collaboration (negotiate, discuss, multiple angles, review from different perspectives, get feedback from multiple experts), task benefits from diverse viewpoints (review, critique, evaluation, brainstorming), or task needs specialist discussion. Examples: 'review this from multiple angles', 'negotiate and handle this issue', 'get different perspectives on this', 'discuss pros and cons of this approach'. |
| user-invocable | false |
Collaboration Skill
Guidance for orchestrating multi-AI collaboration in Unitor.
When to Use Collaboration
Use /unitor:collab when a task requires multiple AI specialists working together:
Tasks needing multiple perspectives:
- Content review (e.g., "Review this marketing copy from multiple angles")
- Design critique (e.g., "Get different perspectives on this API design")
- Architecture discussion (e.g., "Discuss pros/cons of this approach")
- Problem analysis (e.g., "Negotiate and handle this copy issue")
- Brainstorming (e.g., "Generate ideas from different viewpoints")
Tasks spanning multiple technical domains:
- Full-stack features (e.g., "Build user profile with API and React UI")
- End-to-end implementation (e.g., "Implement authentication system")
- Cross-domain integration (e.g., "Create registration with backend and frontend")
Collaboration keywords to watch for:
- Negotiate, discuss, multiple perspectives, multiple angles, different viewpoints
- Review, critique, evaluate, improve, feedback
- Diverse opinions, team discussion, brainstorm
- "Get feedback from", "multiple perspectives on", "different angles"
How It Works
You (Claude) are the coordinator. The workflow:
- Analyze the task - Understand what needs to be built
- Define roles - Determine required specialists (not limited to frontend/backend/database)
- Write role descriptions - Create specific, task-focused descriptions
- Create roles JSON - Write to temporary file
- Execute - Call runtime with roles file
- Report - Present results to user
Role Descriptions
Write task-specific descriptions (not generic templates):
Good:
- "JWT auth API - implement /login, /register, /refresh with jsonwebtoken, bcrypt hashing, token validation middleware"
- "React login UI - build login/signup forms with email/password validation, error display, loading states"
Bad:
- "Backend developer"
- "Frontend specialist"
Example
Task: "Build user authentication"
Roles JSON:
{
"roles": {
"auth-api": "JWT authentication API - implement login/register/refresh endpoints with token generation, validation middleware, bcrypt password hashing",
"login-ui": "React authentication UI - build login/signup forms with validation, error handling, loading states, token storage",
"user-db": "User database - design users table with email, password_hash, created_at, sessions table for refresh tokens"
}
}
Runtime call:
node "${CLAUDE_PLUGIN_ROOT}/scripts/unitor-runtime.mjs" collab /tmp/roles.json "build user authentication"
Your Role
- Analyze each task individually
- Generate specific role descriptions
- No hardcoded templates or rules
- Each task is unique