agent-team-driven-development
About agent-team-driven-development
agent-team-driven-development is a skill for coordinating specialized agents executing an implementation plan in parallel. It helps choose between Agent Teams and sub-agents based on dependencies and communication needs, and covers assignment, coordination, and verification.
Key features
- 01Assess task independence and file conflicts
- 02Assign implementer, reviewer, and architect roles
- 03Coordinate through a shared task list
- 04Require verification evidence
- 05Validate integration and clean up team resources
Use cases
- 01Execute a complex implementation plan in parallel
- 02Coordinate development work that needs cross-stream discussion
Install
Copy the command and run it in a terminal.
npx skills add https:/After installing, type/agent-team-driven-developmentin Claude Code to confirm it loaded.
Bundled files
References and scripts installed along with SKILL.md.
references/architect-role.mdimplementer-role.mdinitiate-team-workflow.mdmanage-team-workflow.mdofficial-documentation.mdreviewer-role.mdDetails
From the project READMEAgent Team Driven Development for Plan Execution
Coordinate multiple specialized teammates working in parallel to execute complex implementation plans.
Agent Teams vs Sub-agents
Choose based on whether workers need to communicate with each other.
| Dimension | Sub-agents | Agent Teams |
|---|---|---|
| Communication | Results return to caller only | Teammates message each other directly |
| Coordination | Main agent manages all work | Shared task list with self-coordination |
| Best for | Focused tasks where only the result matters | Complex work requiring discussion and collaboration |
| Token cost | Lower | Higher: each teammate is a separate instance |
Use Sub-agents for independent tasks needing no inter-worker communication (research, validation, file search). Use Agent Teams when teammates must share findings, challenge each other, or self-coordinate across 3+ parallel work streams. For sequential or highly interdependent tasks, use a single session.
Execution Workflow
- Analyze plan -- identify task independence, file conflicts, and required roles (Implementer, Reviewer, Architect)
- Spawn team -- provide each teammate with task assignments, file paths, constraints, and verification criteria. Teammates do not inherit conversation history.
- Coordinate -- monitor via shared task list, facilitate cross-teammate communication, use delegate mode (
Shift+Tab) to keep the lead focused on coordination - Verify and clean up -- validate integration, run tests, shut down teammates, clean up team resources via the lead
See ./references/initiate-team-workflow.md and ./references/manage-team-workflow.md for detailed workflows.
Roles
- Implementer: executes coding tasks on assigned files, follows TDD/BDD. See
./references/implementer-role.md. - Reviewer: validates quality, security, and plan compliance. See
./references/reviewer-role.md. - Architect: resolves cross-cutting concerns, maintains system-wide consistency. See
./references/architect-role.md.
Key Practices
- Assign distinct file ownership per teammate to prevent edit conflicts
- Include 5-6 tasks per teammate for steady throughput
- Document task dependencies explicitly so blocked tasks wait automatically
- Provide full context in spawn prompts (file paths, goals, constraints)
- Require verification evidence (test results, etc.) upon task completion
- Monitor frequently; unattended teams risk wasted effort
For architecture, capabilities, and limitations, see ./references/official-documentation.md.