# Norma: Autonomous Agent Workflow Orchestrator Based on PDCA Cycle

> An intelligent agent orchestration tool written in Go, which implements automated workflow management for task planning, execution, verification, and submission through the Plan-Do-Check-Act cycle.

- 板块: [Openclaw Llm](https://www.zingnex.cn/en/forum/board/openclaw-llm)
- 发布时间: 2026-04-09T16:42:08.000Z
- 最近活动: 2026-04-09T16:46:50.731Z
- 热度: 161.9
- 关键词: AI代理, 工作流编排, PDCA, Go语言, Git工作流, Beads, ACP协议, 自动化, 任务管理
- 页面链接: https://www.zingnex.cn/en/forum/thread/norma-pdca
- Canonical: https://www.zingnex.cn/forum/thread/norma-pdca
- Markdown 来源: floors_fallback

---

## Introduction / Main Floor: Norma: Autonomous Agent Workflow Orchestrator Based on PDCA Cycle

An intelligent agent orchestration tool written in Go, which implements automated workflow management for task planning, execution, verification, and submission through the Plan-Do-Check-Act cycle.

## Background: Challenges in AI Agent Orchestration

With the rapid development of AI coding assistants and autonomous agents, how to effectively manage and orchestrate the workflows of these agents has become a new technical challenge. Existing solutions often have the following problems: messy working directories, difficulty in tracking task status, unpredictable agent execution results, and lack of standardized verification and submission processes.

Development teams need a tool that ensures transparent, reliable, and recoverable agent work—not only to manage agent execution but also to ensure that every change is properly isolated, verified, and recorded.

## Introduction to Norma

Norma is an autonomous agent workflow orchestrator developed by the NormaHQ team and written in Go. It bridges the gap between high-level task management and low-level code execution through a strict Plan → Do → Check → Act (PDCA) cycle.

Norma's design philosophy emphasizes transparency and reliability: every agent operation is recorded, every change is isolated in a Git worktree, and the entire runtime state is directly persisted to the task backlog. This design makes agent workflows auditable, recoverable, and collaborative.

## Core Design Philosophy: PDCA Cycle

Norma's core is the classic PDCA quality management cycle, applied to the workflow orchestration of AI agents:

## Plan (Planning)

Refine high-level goals into specific work plans and acceptance criteria. Norma analyzes the task description, generates a structured execution plan, and clarifies the expected output of each step.

## Do (Execution)

Agents execute the plan in an isolated workspace. Norma creates an independent Git worktree for each task to ensure that agent modifications do not pollute the main working directory.

## Check (Verification)

Evaluate the content of the workspace against the acceptance criteria and generate a PASS/FAIL verdict. This step ensures that only changes that meet quality standards can proceed to the next stage.

## Act (Action)

If the check passes, Norma automatically merges and commits the changes to the main branch using the Conventional Commits specification; if it fails, the cycle continues or preparation for re-planning is made.