team-assemblelisted

## Purpose Use this skill when fixed engineering-only teams are not enough and the roster should adapt to task shape. ## Trigger - User asks to "do it as a team" - Task mixes domains (for example research + implementation + reporting) - Team composition and model budget must be explicit before execution ## Workflow 1. Analyze task shape (domain, output target, risk posture). 2. Select domain specialists and format specialists separately. 3. Spawn parallel research lanes (`omi-researcher` xN) when breadth is high. 4. Assign decision lane (`omi-consultant` or `omi-architect`) and quality lanes (`omi-reviewer`, `omi-verifier`, `omi-debugger`). 5. Assign model profile by role: - judgment/gating: `glm-5` - execution-heavy: `glm-4.7` - broad low-risk exploration: `glm-4.7` 6. Ask approval gate: "Proceed with this team? (yes/no)". 7. On approval, run the staged pipeline (`team-plan -> team-prd -> team-exec -> team-verify -> team-fix`) with explicit handoffs. ## Output Template ```markdown ## Team Fit Analysis - ... ## Proposed Team | Role | Agent | Why | Model | Lanes | | --- | --- | --- | --- | --- | ## Collaboration Protocol - ... ## Approval Gate - approval detected: - prompt: ## Execution Summary (only when approved) - ... ``` ## Notes - Keep role assignment minimal and explain why each role exists. - Prefer explicit verify/fix triggers over vague quality claims.