teamwork

Coordinates a role-based subagent team for multi-milestone tasks through bounded routing requests.

Updated Jul 12, 2026
One-click install
npx skills add https://github.com/sumitake/agent-collab --skill teamwork-sumitake
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: teamwork
Source: https://github.com/sumitake/agent-collab/tree/main/plugins/agent-collab/skills/teamwork
Command: npx skills add https://github.com/sumitake/agent-collab --skill teamwork-sumitake

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Multi-milestone work often mixes research, code generation, review, and integration in one undifferentiated effort, making ownership, independence of review, and acceptance evidence unclear. This Skill decomposes such work into explorer, worker, reviewer, and integrator roles with explicit budgets, authority, and stop conditions. ## Core Features & Use Cases - Role-based decomposition: Assigns read-only exploration, bounded codegen, review with verified independent lineage, and primary-owned integration to separate subagents. - Bounded routing dispatch: Sends one JSON routing request per work unit through the plugin's coordinator, with dependency ordering, quality/effort selection, and optional explicit reviewer targeting. - Milestone slicing guidance: Applies tracer-bullet vertical slices for product-feature milestones and expand-contract sequencing only for staged migrations that genuinely require coexistence. - Use Case: A user asks to "run this feature build as a team" — the primary plans milestones, dispatches a worker for a vertical slice, routes an independent reviewer with verified lineage, and integrates accepted patches with an attributed ledger. ## Quick Start Ask the agent to run your multi-milestone task as a team, for example by saying "use teamwork to plan and implement this feature with an independent review."

Frequently Asked Questions about teamwork

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run a multi-milestone task as a team of AI subagents?

Invoke the teamwork skill by asking to run the task as a team. The primary agent decomposes work into explorer, worker, reviewer, and integrator roles, then dispatches one bounded routing request per independently useful deliverable with explicit dependencies.

How does the teamwork skill verify an independent code review?

The caller establishes primary, artifact-author, and reviewer lineages before dispatch and verifies the observed reviewer family and source afterward. A routing result alone never proves independence; unverified reviews stay advisory with no independent-approval claim.

When should worker milestones use vertical slices versus expand-contract sequencing?

Product-feature milestones prefer tracer-bullet vertical slices that cut a narrow complete path through all layers. Expand-contract sequencing is reserved for staged migrations where old and new forms must genuinely coexist to keep callers and CI working.

Does the teamwork skill execute provider commands directly?

No. It only sends bounded JSON routing requests to the plugin's coordinator.py and never discovers provider executables, reconstructs raw commands, or retries consumed work units. Provider state remains unknown unless native evidence establishes it.

What are the limitations of planning-only routing requests?

A planning request with dispatch_requested=false reports route eligibility only, not live availability or authentication. Content availability and each work unit's execution status are separate facts that planning cannot establish.