From project-coordinator
This skill should be used when the user asks to "manage this complex task", "track this project", "coordinate this work", "I keep losing track", "investigate this bug", "このタスクを管理して", "進捗を追跡して", "迷子になってきた", "このバグを調査して", or mentions project coordination, purpose tracking, or plan management. Provides orchestration for complex, uncertain tasks while maintaining focus on original objectives.
How this skill is triggered — by the user, by Claude, or both
Slash command
/project-coordinator:project-coordinatorThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Launch project coordination for complex, uncertain tasks.
Launch project coordination for complex, uncertain tasks.
.claude/project-coordinator/ directory exists (ongoing project)NOT for: Predictable, low-risk tasks that existing agents handle well.
.claude/project-coordinator/ for existing docspurpose-extraction skill and apply it to clarify with user.CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS is available)Execute the following steps exactly in order. Do not skip, reorder, or improvise.
⚠️ Critical rules — violating ANY of these breaks the workflow:
TeamCreate(
team_name: "project-coordinator",
description: "<1-line summary from purpose.md>"
)
Create exactly one task. This is the entire project objective, not a subtask.
TaskCreate(
subject: "<user's task in imperative form>",
description: "## Purpose\n<purpose.md content>\n\n## Task\n<user's original request>",
activeForm: "<present continuous form of the task>"
)
Record the returned task ID for Step 4.
Task(
subagent_type: "project-coordinator:coordinator",
name: "coordinator",
team_name: "project-coordinator",
description: "Coordinate project investigation",
prompt: "You are the coordinator for this team.\nYour teammate is \"investigator\" — send investigation tasks via SendMessage.\nCheck TaskList, claim your task, read purpose.md, create plan.md, then begin."
)
Task(
subagent_type: "project-coordinator:investigator",
name: "investigator",
team_name: "project-coordinator",
description: "Investigate on coordinator direction",
prompt: "You are the investigator for this team.\nWait for coordinator to send you tasks via messages.\nDo NOT start working until coordinator gives you a specific task."
)
TaskUpdate(
taskId: <task ID from Step 2>,
owner: "coordinator"
)
Do NOT set status: "in_progress" — coordinator does this when it claims the task.
After Step 4, the lead's launch work is complete.
The lead should wait for coordinator's reports. Do not poll or micromanage. User can message either teammate directly (Shift+Up/Down in terminal).
Read ${CLAUDE_PLUGIN_ROOT}/agents/coordinator.md and apply its principles in this session.
For investigation tasks, call investigator via Task tool:
Task tool:
subagent_type: "project-coordinator:investigator"
prompt: "## Context\n[Purpose summary]\n\n## Step\n[Current step]\n\n## Task\n[Investigation details]"
⚠️ CRITICAL: Never just mention "delegate to investigator". Always use Task tool explicitly.
npx claudepluginhub suzuki0keiichi/claude-plugins-suzuki0keiichi --plugin project-coordinatorSplits large tasks into research, synthesis, implementation, and verification phases, assigning ownership to bounded workers while the coordinator focuses on integration.
Decomposes complex work into trackable tasks with dependency chains using TaskCreate, TaskUpdate, TaskGet, TaskList tools. For multi-step implementations, parallel coordination, and status tracking.
Orchestrates multi-agent parallel execution for complex tasks like features, refactoring, testing, reviews, and documentation using cc-mirror tracking and TodoWrite visibility.