Can I prevent subagents from making unauthorized changes?

Yes, you can spawn teammates with a 'Require plan approval before making changes' flag, allowing the lead agent or user to review their strategy first.

Do subagents inherit my full conversation history?

No, subagents do not inherit the lead's history. Instead, the lead provides a specific 'Spawn Context' with relevant files and constraints to keep the subagent focused and efficient.

How does Team Plan handle file conflicts?

The workflow includes a 'File Ownership' step during the planning phase, ensuring that no two teammates are assigned to write to the same file simultaneously.

How does the scouting phase improve agent performance?

The scouting phase uses parallel subagents to gather context before any code is changed. This ensures the plan is based on the actual codebase state and current documentation rather than assumptions.

What is the benefit of EnterPlanMode?

EnterPlanMode allows the agent to pause execution and think through the strategy, team roles, and dependencies, presenting a clear roadmap for user approval before work begins.

Team Plan & Agent Orchestrator

Name: Team Plan & Agent Orchestrator
Author: zhangxin840

byzhangxin840

•

生産性とワークフロー

Orchestrates complex tasks using a scout-then-plan workflow to coordinate multiple specialized subagents and domain researchers.

The Team Plan skill introduces a sophisticated multi-agent orchestration pattern for Claude Code, designed to handle high-complexity engineering tasks. It operates in two distinct phases: a parallel 'Scout' phase where subagents explore the codebase and research external documentation to gather deep context, followed by a structured 'Plan' phase. During planning, the skill defines specific team roles, task dependencies, and file ownership to ensure that large-scale implementations are executed with precision and without conflicting edits between subagents.

主な機能

01Granular task breakdown with dependency tracking (blockedBy)

02Structured Plan Mode for task analysis, risk assessment, and team design

03Parallel context gathering via specialized Explore and Researcher agents

04Customizable subagent spawn context to optimize token usage and focus

051 GitHub stars

06Explicit file ownership to prevent write conflicts between subagents

ユースケース

01Implementing complex new features requiring deep research into third-party APIs and local patterns

02Managing high-risk changes by requiring plan approval for specific subagent actions

03Executing large-scale architectural refactors across multiple directories

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add zhangxin840/team-plan-skill team-plan

For use in Claude.ai and ChatGPT

Download Skill