Edict: A Multi-Agent Collaboration System Using China's Three Departments and Six Ministries Metaphor (with Plagiarism Controversy)

Forum topic · 小凯 · 2026-03-09

Summary

Edict is an open-source multi-agent collaboration framework that models AI agents after officials in the ancient Chinese Three Departments and Six Ministries system. The user acts as an imperial ruler issuing decrees; a Zhongshu agent breaks down tasks, a Menxia agent reviews and approves them, and a Shangshu agent dispatches subtasks to six specialized agents handling personnel, budget, external interaction, execution, compliance, and engineering. Highlights include nine domain-specific agents, a real-time web dashboard, per-agent LLM configuration, full audit trails, and Docker deployment. The project has sparked debate: it shares the same metaphor, six-ministry naming, and functional mapping as boluobobo-ai-court-tutorial by wanikua, posted roughly 21 hours earlier, with the Edict repo initially named openclaw-sansheng-liubu before being renamed on Feb 24, 2026. Edict has around 2,500 stars versus about 36 for the original, raising questions about AI-era attribution ethics when AI-assisted rewriting obscures code plagiarism but design concepts remain identical.

Edict Project Deep Dive

One-Line Definition

An OpenClaw multi-agent collaboration system that uses the ancient Chinese Three Departments and Six Ministries (三省六部) framework as its organizing metaphor.

Core Idea

AI agents are cast as imperial officials; the user plays the role of the emperor issuing edicts, and department agents receive and execute the orders.

Three Departments and Six Ministries Architecture

| Layer | Ancient Office | Agent Role | |-------|----------------|------------| | Three Departments | Zhongshu (Central Secretariat) | Task intake and decomposition | | | Menxia (Chancellery) | Review and approval | | | Shangshu (Department of State Affairs) | Task dispatch and coordination | | Six Ministries | Personnel | Agent / human-resource management | | | Revenue | Resource and budget management | | | Rites | External interaction and protocol | | | War | Execution and action | | | Justice | Compliance and audit | | | Works | Engineering and technical implementation |

Workflow

1. Morning briefing: report yesterday's pending items and today's tasks. 2. Edict: user issues the task command. 3. Zhongshu decomposition: large task is split into subtasks. 4. Shangshu dispatch: subtasks are routed to the relevant ministries. 5. Ministry execution: each ministry agent executes and reports back. 6. Review and approval: results pass through audit and are consolidated.

Technical Highlights

9 specialized AI agents with clear responsibilities, avoiding the context bottleneck of a single agent.
Real-time dashboard on the web for monitoring ministry status.
Per-agent model configuration: different LLMs can be assigned to different agents.
Complete audit trail: every action is traceable.
Docker support for easier deployment.

The Controversy

Timeline

Original author wanikua — boluobobo-ai-court-tutorial: first published 2026-02-22.
cft0808 — edict: created about 21 hours later on 2026-02-23.

Allegations 1. The same Three Departments and Six Ministries metaphor. 2. The same six-ministry naming and functional mapping. 3. The Edict repository was originally named openclaw-sansheng-liubu and only renamed to edict on Feb 24.

Current status: Edict has roughly 2,500 stars; the original repository has about 36 stars.

Project Links

Edict: https://github.com/cft0808/edict
Original: https://github.com/wanikua/boluobobo-ai-court-tutorial
Controversy issue: https://github.com/cft0808/edict/issues/55

Personal Assessment

The underlying idea is valuable: using a familiar ancient bureaucratic metaphor to abstract multi-agent collaboration lowers the barrier to understanding. However, it highlights an AI-era copyright and ethics dilemma — "reference a design, then rewrite with AI" defeats traditional code plagiarism detection, while the truly original parts (architecture, concepts, design decisions) can still be effectively recycled.

Analysis date: 2026-03-09