# myflow: A Structured 5-Stage Workflow for AI-Assisted Software Development

> myflow is a five-stage pipeline workflow for AI-assisted software development, covering from requirement discovery to code deployment. Each stage has clear deliverables and handoff mechanisms. It integrates the two skill systems of Superpowers and RPIV, providing developers with a complete and unambiguous development methodology.

- 板块: [Openclaw Llm](https://www.zingnex.cn/en/forum/board/openclaw-llm)
- 发布时间: 2026-06-14T02:46:26.000Z
- 最近活动: 2026-06-14T02:48:58.800Z
- 热度: 146.0
- 关键词: AI辅助开发, 软件开发工作流, Pi Agent, RPIV, Superpowers, 结构化开发流程, 代码审查, TDD, 项目管理, 知识管理
- 页面链接: https://www.zingnex.cn/en/forum/thread/myflow-ai
- Canonical: https://www.zingnex.cn/forum/thread/myflow-ai
- Markdown 来源: floors_fallback

---

## myflow: A Structured 5-Stage Workflow for AI-Assisted Software Development

myflow is a five-stage pipeline workflow for AI-assisted software development, covering from requirement discovery to code deployment. Each stage has clear deliverables and handoff mechanisms. It integrates the Superpowers and RPIV skill systems, providing developers with a complete, unambiguous development methodology.

## Background: The Process Dilemma in AI-Assisted Development

With the widespread application of large language models in software development, developers realize that having powerful AI tools alone is insufficient—key lies in integrating these tools into a systematic workflow. Many teams face issues like chaotic patterns, unclear handoffs, and difficulty in沉淀 learning when using AI for programming. myflow was created to address these problems; it's not just a tool set but a complete, structured five-stage pipeline offering clear paths and predictable deliverables for AI-assisted development.

## Core Method: The Five-Stage Pipeline

myflow's core is a five-stage development pipeline covering the full lifecycle: Discover & Align → Research & Design → Implement → Validate & Review → Land & Learn.
- **Discover & Align**: Clarify problems, define scope, reach consensus; key deliverable is Functional Requirement Document (FRD).
- **Research & Design**: Design technical solutions; deliverable is detailed technical plan; uses epiphany-tabling to store ideas without interrupting workflow.
- **Implement**: Write code with TDD, subagents, and verification-before-completion (evidence before claiming completion).
- **Validate & Review**: Ensure code meets expectations and quality standards; includes code review and revision.
- **Land & Learn**: Structured 9-step收尾 with as-built documentation (recording actual delivery) and learning capture.

## Cross-Stage Mechanisms & Knowledge Management

**Cross-stage mechanisms**:
- Epiphany Tabling: Runs in stages 2-4, captures inspiration without interrupting work (stored in docs/tabled.md).
- Learning Capture: Conducted after stages1,2,4,5; promotes observations to persistent knowledge if appearing twice.
- Handoff Management: Uses create-handoff and resume-handoff to pause/resume work at any stage.
**Knowledge management**: myflow defines standard doc structure: docs/tabled.md (stored ideas), docs/status.md (dynamic state), docs/memory/ (persistent memory), docs/changes/ (as-built docs), docs/retros/ (retrospectives), docs/runbooks/ (multi-skill processes), AGENTS.md (agent guidance).

## Upstream Dependencies & Practical Usage

**Upstream dependencies**: myflow integrates two upstream projects—Superpowers (brainstorming, TDD, subagents) and RPIV (discover-blueprint-implement-validate-review framework). It uses a "observe but not auto-integrate" strategy: upstream evolves independently; sync-upstream skill shows changes but doesn't auto-merge.
**Practical steps**:
1. Clone repo and install: `git clone https://github.com/don-smith/myflow.git && pi install ./myflow`
2. Run `/skill:setup-myflow` to initialize repo conventions.
3. Start with `/skill:myflow` → stage1 (use brainstorming or /skill:discover).
4. Finish with `/skill:land`.
5. Sync upstream every 2-3 projects via `/skill:sync-upstream`.

## Conclusion & Insights

myflow's value lies in solving core AI-assisted dev problems:
1. Eliminates ambiguity via clear stages and deliverables—developers always know their stage and next steps.
2. Emphasizes knowledge persistence: as-built docs, retrospectives, learning capture ensure team experience is preserved.
3. Modular design allows gradual adoption: teams can start with single skills and expand to full workflow.
For teams exploring AI-assisted dev best practices, myflow provides a well-thought reference implementation worth studying.