skills.rest
agentscope-ai
agentscope-aiOFFICIAL
Explore all skills by this author →
𝕏@agentscope_aiGitHub

teamharness-project-management

Maintains durable TeamHarness project state via projectflow DAG and Loop planning actions.

5.5k|677|Updated Feb 21, 2026
Search Tags:#task orchestration#Matrix#Multi-Agent#Project management#loop iteration#dag planning#projectflow
One-click install
npx skills add https://github.com/agentscope-ai/HiClaw --skill teamharness-project-management
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: teamharness-project-management
Source: https://github.com/agentscope-ai/HiClaw/tree/main/plugins/teamharness/skills/team/project-management
Command: npx skills add https://github.com/agentscope-ai/HiClaw --skill teamharness-project-management

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Multi-agent project work loses track of plans, dependencies, and accepted progress when state lives only in chat messages. This Skill gives a Leader agent a durable, file-backed project state (meta.json, plan.md, result.md) so task dependencies, Loop iterations, and requester reports stay consistent across sessions.

Core Features & Use Cases

  • Project and Quick Project Creation: Create full projects or single-task quick projects with safe IDs, requester routes, and CoPaw-compatible runtime files under shared/projects/.
  • DAG and Loop Planning: Plan finite dependency graphs with plan_dag, or iterative work with plan_loop, record_loop_iteration, and ready-node resolution, rejecting cycles and duplicate task IDs.
  • Result Acceptance and Reporting: Accept checked Worker results with accept_task_result, resolve project context from a task ID, and generate localized DAG/Loop status reports delivered to the requester.
  • Use Case: A Leader receives a DingTalk request, creates a project in a Matrix task room, plans a three-task DAG across two Workers, accepts each verified result, and sends the requester a progress report with published artifacts.

Quick Start

Ask the Leader agent to create a TeamHarness project for your request, plan the task DAG, and report progress back to you as Worker results are accepted.

Frequently Asked Questions about teamharness-project-management

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

FAQPage Schema
How do I create a project with the projectflow tool?▼

Call projectflow with action create_project and a JSON payload containing title, source, requester, and optionally projectId and replyRoute. This writes meta.json and plan.md under shared/projects/{project-id}. For single-task work, use create_quick_project from the Matrix task room instead.

When should I use plan_dag versus plan_loop?▼

Use plan_dag when the work is finite and dependencies are known upfront; it returns readyNodes you can delegate immediately. Use plan_loop when work must repeat until a stop condition, quality gate, or maxIterations, recording each round with record_loop_iteration.

How does a Leader accept a Worker task result?▼

After task delegation checks return effective: true, call accept_task_result with projectId, taskId, resultStatus, and a summary. This updates the DAG or Loop node, sets requester_report.pending, and unblocks downstream ready nodes.

Can I create a project directly in the requester chat session?▼

No. For Project Work in a dedicated task room, send a PROJECT_REQUESTED self-trigger to the task room first, then call create_project there while preserving the original source, requester, sourceRoomId, and replyRoute. Do not guess route values.

What projectflow actions are not supported?▼

TeamHarness v0.1 supports project creation, quick projects, context resolution, DAG and Loop planning, ready-node resolution, iteration recording, result acceptance, report clearing, pause, resume, and complete. Calling other projectflow actions is rejected, and plans reject duplicate IDs, unknown dependencies, and cycles.

Why does my projectflow call fail with a payload error?▼

The payload must be passed as a JSON object, not a serialized JSON string. Also verify IDs match the safe pattern [A-Za-z0-9][A-Za-z0-9._-]* with no colons, slashes, spaces, or Matrix IDs.

Related Skills

Explore More Product & Management Skills→
joshka0joshka0
3

foxctl-room

Coordinate durable multi-agent room messaging and task workflows with shared state.

Community
Advanced
agentscope-aiagentscope-ai
5.3k

communication

Route and format Leader-to-Worker notifications across Matrix rooms.

Official
Intermediate
ReinaMacCredyReinaMacCredy
223

beads

Track multi-session work with dependency graphs and git-backed persistence.

Community
Advanced
BuidlGuidlBuidlGuidl
12

bgipfs Upload

Upload files, directories, URLs, JSON, and text to IPFS via CLI or SDK.

Official
Intermediate