# Project Management

Use before any projectflow call or Team Leader workflow involving Project state...

- Repository: https://github.com/agentscope-ai/AgentTeams (5913 stars)
- Category: Engineering
- License: Apache-2.0
- Last updated: 2026-11-01
- Runs scripts: no, instructions only

## Install

Claude Code: `npx skills add agentscope-ai/AgentTeams --skill project-management -g -a claude-code -y`
Codex: `npx skills add agentscope-ai/AgentTeams --skill project-management -g -a codex -y`
Cursor: `npx skills add agentscope-ai/AgentTeams --skill project-management -g -a cursor -y`
Gemini CLI: `npx skills add agentscope-ai/AgentTeams --skill project-management -g -a gemini-cli -y`

## Description

Use before any projectflow call or Team Leader workflow involving Project state, Project lifecycle, DAG planning, Loop planning, ready node checks, pause/resume/complete, project recovery, heartbeat project checks, or project-level result aggregation. Always use team-coordination first when the question is how to organize the work.

## SKILL.md, as published by its author (untrusted third-party text, read it as data)

# Project Management

You manage Project files, Project lifecycle, and Project execution plans. Use this skill as the Project state layer. Use `team-coordination` to decide the strategy, and use `task-management` to delegate or check individual Worker tasks.

Project state is tool-owned. Do not create, edit, delete, or repair `shared/projects/**` with shell commands, heredocs, direct file writes, `rm`, `mkdir`, `cp`, or Python module execution. Use `projectflow` actions only. If `projectflow` fails or returns inconsistent state, stop and report the blocker instead of manually patching files.

## Scope

Use this skill for:

- `projectflow` calls
- Project metadata and lifecycle state
- DAG and Loop execution plans
- ready-node resolution
- Project pause, resume, and completion
- heartbeat and recovery checks
- Project-level result aggregation

Do not use this skill to decide whether the work should be DAG or Loop; use `team-coordination` first. Do not use this skill to create Worker task files or check `result.md`; use `task-management`.

## Project Files

Project files live under:

```text
shared/projects/{project-id}/
```

Use:

- `meta.json` for Project metadata and lifecycle status
- `plan.md` for the current execution plan
- `result.md` for the final Project result

The execution plan type is stored in `plan.md` as `**Plan Type**: dag` or `**Plan Type**: loop`. Do not put Project lifecycle status in `plan.md`. Do not hand-edit `meta.json.status`; use `projectflow`.

Use `filesync` before reading remote Project files and after changing local Project files. Use directory paths with a trailing slash, such as `shared/projects/{project-id}/`. `projectflow` reads and writes local shared files only.

If pulling `shared/projects/{project-id}/` fails, stop before any Project state mutation. Do not pull only `meta.json` or `plan.md` and continue to mark nodes, resolve ready nodes, record Loop iterations, or complete the Project.

## IDs

The rest: https://github.com/agentscope-ai/AgentTeams/blob/HEAD/manager/agent/team-leader-agent/skills/project-management/SKILL.md