🔬 Lean AI Methodology · 史凯 Kai Shi · v1.3
Lean AI PRD Team sky791016/lean-ai-dev-team

Scenario-Driven. Value-First. Minimum Viable MVP, Every Time.

A structured prompt system that routes your task through a 8-agent coordinated team — locking in the business scenario and ROI before writing a line of code, delivering the smallest valuable loop that proves value, and cutting token waste by replacing open-ended chat with structured phase handoffs. Works with Claude Code, Cursor, Windsurf, JetBrains, 通义灵码, CodeBuddy and more.

Install the Skill → Meet the Team

精益AI方法论（Lean AI Methodology） — 史凯 Kai Shi

"AI transformation is not model procurement — it is lean reconstruction of processes, data, organization, technology, and operations, with business scenarios as the core."
Scenario First Value-Driven Small Steps, Fast Cycles Data Before Code Controlled Execution Continuous Operations Human-AI Collaboration
8
Specialized AI Agents
12+
Compatible IDEs
4
Closed-Loop Checks
0
Blind Code Stubs

The Core Problem

Large Models Generate Waste.
Lean AI Eliminates It.

In the AI coding era, the bottleneck is no longer writing code — it's knowing which code to write, for whom, and why. Without a value-native method, AI amplifies waste, not value.

⚠ The 7 New Wastes of AI-Era Development

W1
Scenario-less Code
AI generates technically correct code for the wrong problem. No scenario definition = guaranteed misalignment.
W2
Metric-free Features
Features ship without KPIs. No one measures adoption, accuracy, or business impact. AI helps you ship faster into a void.
W3
Hallucinated Architecture
AI invents API endpoints, modules, and integrations that don't match the existing system. No one reads the code first.
W4
Data Loop Neglect
AI systems launch but never improve. No feedback loop means the model stays static while the world changes.
W5
Governance Debt
AI agents act without audit trails, permission checks, or human escalation gates. Risk compounds silently.
W6
Platform Before Scenario
Teams build AI platforms first, then search for use cases. The platform becomes shelfware before anyone uses it.
W7
Trust Erosion
AI makes one critical mistake. Without controlled execution design, the team abandons AI entirely — the worst waste of all.

❌ Without Lean AI Method

→ Prompt → Code → Ship → Hope it works
→ No ROI model. No KPIs. No feedback loop.
→ Frontend calls APIs that don't exist yet
→ AI writes for a system it never read
→ 6 months later: "the AI project failed"
→ Core systems quietly broken by AI overwrites
→ Users don't trust the output; adoption <20%

✅ With Lean AI PRD Team

→ Scenario → Value → Architecture → Code → Measure
→ ROI model before line one of code is written
→ API contracts signed by Architect before build
→ Every agent reads existing code first
→ Four closed loops checked before DoD is signed
→ Clean Core: core systems untouched by AI agents
→ Controlled execution gates prevent trust erosion

The Lean AI Insight

"In the AI coding era, the real bottleneck is not generation speed — it's value clarity. Large models make it trivially easy to generate code. They make it no easier to generate the right code for the right problem."

Lean AI PRD Team solves this by embedding the Lean AI Methodology directly into the agent workflow. Every invocation forces scenario definition, ROI measurement, architecture contracts, and four-loop closure — before a single line of production code is written.

14×
faster ROI validation
vs unstructured AI coding
0
blind code stubs generated
agents read code first
4
closed loops verified
before every DoD sign-off
8
specialized agents
each owns a distinct domain

— 精益AI方法论 · Kai Shi (史凯) · sky.kugua@gmail.com · Founder of Lean AI Method

Live Demo

Watch the Team Activate

Type one command. Eight agents coordinate automatically — scenario and value locked in first, code last.

claude-code — ~/my-project
$ /lean-ai-prd-team Build a procurement AI agent for contract risk review
⚡ Lean AI PRD Team activated — 8 agents loading...
▶ Phase 1 · 业务规划师 ........................................ ✓ done
Scenario Level: L3 (流程协同) · Value Type: 控险 + 增效
Stakeholders: Legal team, Procurement ops, CFO · Roadmap: POC 6wk → MVP 3mo → Scale Q4
▶ Phase 2 · 产品经理 ........................................ ✓ done
ROI: 合同审查时间 −70% · 人力节省 ¥2.4M/yr · 风险识别率 +55%
KPIs: 采纳率 >80% · 准确率 >92% · 人工接管率 <15%
▶ Phase 3 · 业务分析师 ......................................... ✓ done
8 user stories · As-Is→To-Be flow · 12 acceptance criteria
Human-AI split: AI flags risk clauses, human approves final verdict
▶ Phase 4 · 技术架构师 ......................................... ✓ done
Clean Core: ERP untouched · Cognitive Sidecar: RAG + LLM review layer
5 ADRs · REST API contracts (11 endpoints) · Governance: L2 controlled exec
▶ Phase 5 · Parallel build [Frontend + Backend + Data] ✓ done
14 files changed · 3 new API endpoints · Migration SQL ready · Knowledge base schema defined
▶ Phase 6 · 合规PM ............................................. ✓ done
✅ Value Loop · ✅ Data Loop · ✅ Model Loop · ✅ Ops Loop
0 interface conflicts · DoD signed · Execution checklist: 23 steps
✨ Report ready. All 8 agents completed. Total wall time: ~4 min.
█

Why This Skill

Five Guarantees. Zero Wasted Tokens.

LLM hallucinations, scope creep, disconnected stubs, runaway token costs — these aren't model problems. They're structure problems. Solved by forcing scenario clarity, value alignment, and minimum MVP scope before a single line of code is written.

🎯

Scenario-Driven, Value-First

Every task is anchored to a business scenario and ROI model before any agent writes code. The Business Planner and PM lock in the "why" and the metric — eliminating scenarioless code, the #1 AI waste.

🏁

Minimum Viable MVP — Not a Feature Dump

Business case → requirements → architecture → parallel code → 4-loop sign-off. The team asks "is this MVP necessary?" at every step. You get the smallest loop that proves value — not a pile of unshippable stubs.

⚡

Saves Tokens — Structured Handoffs

Structured phase handoffs replace open-ended chat. No iterative clarification loops. One structured prompt replaces dozens of back-and-forth messages — you get all 9 roles in one focused pass, with zero scope-creep overhead.

🏗️

Read Before You Write

Every agent reads existing code before generating anything. No phantom modules, no mismatched APIs — the team works with your actual codebase, not a blank canvas.

⚖️

Four Closed-Loop Guarantee

Value loop, data loop, model loop, and operations loop are checked at the end of every task. Nothing ships without all four confirmed — ensuring value realization, not just feature delivery.

🛡️

Works With Any IDE

The skill is structured markdown — it works with Claude Code natively, or paste SKILL.md into Cursor, Windsurf, JetBrains, 通义灵码, CodeBuddy, Comate, or any LLM API.

Real Use Cases

From One Command to Production-Ready Systems

Three enterprise teams used Lean AI PRD Team to ship AI-native features — with measurable ROI on day one.

CASE 01
Enterprise Procurement · L3 流程协同

Contract Risk AI Agent

A manufacturing company needed AI to review 500+ contracts per month for risk clauses, compliance violations, and unfavorable terms — without changing their SAP core.

Contract review time −72%
Risk clause detection rate +58%
Annual labor savings ¥2.4M
MVP shipped in 11 days
// 产品经理 output · Scenario Card
场景名称: 合同条款风险审查智能体
业务目标: 控险 + 增效
场景级别: L3 流程协同
人机分工:
AI: 条款识别, 风险标注, 合规检查
人工: 最终判断, 高风险条款审批
价值指标:
审查时间: 4h → 40min (-83%)
识别准确率目标: >92%
ROI回收期: 4.2个月
// 技术架构师 · ADR-001
决策: Clean Core + Cognitive Sidecar
SAP untouched · RAG on contract DB
Review layer via API · Audit log: required
// 技术架构师 · System Design
四层架构:
L1 交互层: WeChat Work + Web UI
L2 认知层: Intent → RAG → Qwen3
L3 行动层: 工单API · 权限闸门
L4 核心层: CRM + 工单系统(只读)
API Contracts (6 endpoints):
POST /api/complaint/classify
POST /api/complaint/draft-reply
GET /api/complaint/:id/history
POST /api/complaint/escalate
GET /api/kb/search
POST /api/complaint/close
// 四闭环 · Compliance PM
✅ Value ✅ Data ✅ Model ✅ Ops
0 conflicts · DoD: 18 items signed
CASE 02
Customer Service · L2 岗位助手

Customer Complaint AI Assistant

A retail platform handling 3,000+ daily complaints needed an AI agent to classify, draft responses, and escalate — integrated with their existing CRM without modification.

Average response time 8min → 90sec
Agent adoption rate 91%
CSAT improvement +23 pts
Time to MVP 8 days
CASE 03
HR Tech · L2 岗位助手

Resume Screening & Interview AI

An enterprise HR team screening 10,000 monthly applicants needed AI to score resumes, generate structured interview guides, and flag compliance risks — all auditable.

Screening time per candidate 45min → 3min
Quality-of-hire improvement +31%
HR team capacity freed 60%
Time to MVP 14 days
// 产品经理 · ROI Model
投入测算:
开发: 14人天 × ¥3,000 = ¥42,000
月运营: Token ¥800 + 维护 ¥2,000
产出测算:
节省工时: 8,500 hr/mo × ¥120
= ¥1,020,000/月
质量提升: 降低错误招聘成本估算
= ¥180,000/月
ROI回收期: < 2周
年收益: ¥14.4M+
// 合规PM · Compliance flags
⚠ GDPR: candidate data minimization
⚠ Bias audit: quarterly required
✅ Audit log: every AI decision logged
✅ Human override: always available

Interactive Demo

Live Agent Workflow — Click to Run

Press ▶ Run Demo to watch all 8 agents execute a real contract risk AI project — scenario locked first, minimum MVP delivered last.

claude-code — ~/enterprise-ai-project
📋 Command
🧭 P1 · Planner
📊 P2 · PM
📋 P3 · Analyst
🏗 P4 · Architect
⚡ P5 · Build
✅ P6 · Closure
# Type your task and press ▶ Run Demo above

$ █
Ready — press ▶ Run Demo to start Lean AI PRD Team · 8 agents · Apache 2.0 · Kai Shi

How to Use · 三种使用场景

One Skill. Three Scenarios. Zero Guesswork.

Pick your scenario, paste the prompt, watch 8 agents coordinate. Every scenario ships a full report — strategy to code to compliance sign-off.

全新项目 — 从一句话需求到完整交付

适用：新功能、新产品、AI 系统从零构建。9 个智能体全部启动，从战略到代码一步到位。

提示词示例 · Prompt
/lean-ai-prd-team [全新项目] 项目背景：法务团队每天审查 50+ 份合同， 平均耗时 2 小时/份 目标：构建合同风险审查 AI 智能体， 自动识别高风险条款并给出修改建议 约束：核心 ERP 不能修改，需要人工最终确认 技术栈：Python Flask + PostgreSQL + React
全部 8 智能体 ROI 先行 API 契约锁定
团队输出预览 · Output Preview
▶ Phase 1 · 业务规划师 ✓
L3 流程协同 · 控险+增效 · 3阶段路线图
▶ Phase 2 · 产品经理 ✓
ROI: −70%审查时间 · ¥2.4M/年节省
▶ Phase 3 · 业务分析师 ✓
8 用户故事 · 12 条 Given/When/Then
▶ Phase 4 · 技术架构师 ✓
5 ADRs · 11 REST API 契约 · Clean Core
▶ Phase 5 · 前端+后端+数据集成 ✓
14 文件变更 · 迁移 SQL · 知识库 schema
▶ Phase 6 · 合规PM ✓
✅ 四闭环通过 · 23步执行清单 · DoD签核

The 8-Agent Team

Scenario to Minimum MVP, Without Gaps

Each agent owns a distinct domain. Earlier agents enforce value clarity and MVP scope constraints that later agents must honor — especially scenario cards, ROI models, and API contracts.

Phase 1 · Strategy
业务规划师 / Business Planner
Sets the strategic frame before any analysis begins. Defines the business case, scenario level (L1–L5), value type (降本/增效/控险), stakeholder map, and 3-stage roadmap. Asks "is AI actually necessary?" before proceeding.
Business Case L1–L5 Scenario Level 3-Phase Roadmap
Phase 2 · Product
产品经理 / Product Manager
Owns value quantification and MVP scope. Produces the Lean AI scenario card, quantified ROI model, minimum viable success KPIs, stop conditions, and go/no-go criteria. Removes features that don't move the needle.
Scenario Card ROI Model KPI Dashboard Stop Conditions
Phase 3 · Analysis
业务分析师 / Business Analyst
Translates MVP scope into engineering specs. User stories scoped to the minimum viable loop, As-Is → To-Be flows, Given/When/Then acceptance criteria, and human-AI handoff design. Defers non-MVP work to backlog.
User Stories As-Is / To-Be Acceptance Criteria Human-AI Split
Phase 4 · Architecture
技术架构师 / Technical Architect
Designs the system before any code is written. ADRs, Clean Core + Cognitive Sidecar design, full API contracts, governance framework, data flows.
ADRs API Contracts Clean Core Design Governance Plan
Phase 5 · Build (Parallel)
前端开发 / Frontend
Implements UI following the architect's API contracts. Reads existing components first, builds human-AI collaboration interfaces, exposes operational dashboards.
Component Changes API Usage Map Ops Dashboard UI
Phase 5 · Build (Parallel)
后端开发 / Backend
Implements APIs per the architect's contracts. Follows existing service patterns, adds @Transactional for multi-table writes, logs all controlled actions.
Endpoints List Service Logic Audit Logs
Phase 5 · Build (Parallel)
数据集成 / Data Integration
Handles schema, migrations, knowledge base design, and external APIs. Crucially: designs the data feedback loop so AI outputs improve future models.
Migration SQL Data Flow Feedback Loop Knowledge Base
Phase 6 · Closure
合规项目管理 / Compliance PM
Consolidates all outputs, runs the four-loop check, detects interface conflicts, produces ordered execution checklist and Definition of Done.
4-Loop Check Conflict Report Execution Checklist DoD
Phase 0 · Audit (重构/评审必跑)
代码审计师 / Code Auditor
重构优化必跑 项目评审必跑
Audits existing code before any refactor or review begins. Detects security vulnerabilities (OWASP Top 10), performance bottlenecks (N+1 queries, missing indexes), code quality issues, and architecture debt. Produces a graded issue list so the team knows exactly what to fix first.
Security Vulnerabilities Performance Bottlenecks Architecture Health Score Prioritized Fix List

Lean AI Framework

Built on 一心·两翼·三层·四闭环

Every agent prompt is aligned to Kai Shi's Lean AI Framework — the same methodology used for enterprise AI transformation at scale.

💰
Value Loop
Every task answers: what business problem? for whom? measurable how? worth scaling?
🗄️
Data Loop
AI outputs, user corrections, and failures feed back into knowledge bases and training sets.
🤖
Model Loop
Architecture decisions support future model swaps. No lock-in to a single provider.
📈
Operations Loop
KPI dashboards, usage metrics, and prompt optimization cadence are defined before launch.

Execution Flow

How the 6 phases connect

Phase
1

业务规划师 — Strategic Frame

Sets scenario level (L1–L5), value type, stakeholder map, and 3-phase roadmap. Nothing begins until "why" is answered.

Phase
2

产品经理 — Scenario Card & ROI

Produces the standard Lean AI scenario card, quantified ROI model, KPI dashboard template, and stop conditions.

Phase
3

业务分析师 — Engineering Specs

User stories, process maps (As-Is → To-Be), acceptance criteria, human-AI handoff nodes, and data requirements.

Phase
4

技术架构师 — System Blueprint

ADRs, Clean Core + Cognitive Sidecar design, full API contracts, governance framework. Frontend and backend must honor this.

Phase
5

前端 + 后端 + 数据集成 PARALLEL

Three agents build simultaneously. All three work from the architect's API contracts and read existing code before writing.

Phase
6

合规项目管理 — Four-Loop Closure

Consolidates all outputs. Runs value, data, model, and ops loop checks. Flags interface conflicts. Signs off on DoD.

Release Notes

What's New

Version history for Lean AI PRD Team Standard & Pro.

v1.3 Interactivity Layer 2026-05
Standard + Pro
💬
Skill Activation Greeting
Every invocation opens with Kai Shi's team introduction and core philosophy — before any analysis begins.
🧑‍💼
Agent Self-Introductions
Each agent opens with a first-person introduction — role, philosophy, and a Kai Shi quote at maximum-impact moments.
📊
Key Findings Output
Every agent surfaces a structured summary to the user before writing the phase file. No need to open .dev-team/ to track progress.
🚦
Clarification Gates
Business Planner, BA, and Architect ask the user one focused question when encountering key ambiguities — never silently guess.
🔁
Phase Handoff Messages
Each agent ends with a one-line handoff naming the next agent — creating a continuous team-collaboration feel.
⚖️
Compliance PM Verdict Card
ASCII-bordered Go / Conditional Go / No-Go verdict with 4-loop results and Kai Shi attribution — every delivery closes formally.
v1.2 Methodology Embedding + English Translation + Dual Persona 2026-05
Standard + Pro
✦ Lean AI Foundation (L1–L5, W1–W7, 4 Loops, Clean Core) embedded in all agents as prerequisite
✦ Full English translation of all agent prompts, steps, and deliverables
✦ Phase 0.5 Idea Intake for first-time users with vague input
✦ [精益诊断] Waste Diagnosis scenario — W1–W7 walkthrough → minimum effective phase routing PRO
✦ Project scaffolding for new projects in Phase 5 agents
✦ Final Report Section 13: "Next Steps for Non-Technical Owners"
v1.1 Structural Fixes + Orchestration 2026-05
Standard + Pro
✦ Added Orchestration section — Claude Code now runs as true multi-agent (not merged single-agent)
✦ Phase 5 split into 5a (Frontend/Backend/Data, parallel) and 5b (Value Assessor, sequential) PRO
✦ Gate Rule added — verify input files before each phase, never fabricate prior content
✦ Fixed routing tables for [重构优化] — broken dependency chains resolved in both versions
✦ Final Report Template wired to Phase 6 Compliance PM (was previously a dead template)
v1.0 Initial Release 2026-04
First public release. Standard (8 agents) and Pro (10 agents). 5 scenarios: [全新项目] · [重构优化] · [项目评审] · [修复] · [前端].

Get Started

Works With Any IDE

The skill is structured markdown — paste it into any AI coding tool. Native slash command for Claude Code.

Standard
/lean-ai-prd-team
✅ 8 agents
✅ Business Planner → PM → BA → Architect → Build → PM
✅ Idea Intake for first-time users
✅ 4-loop closure check
✅ Free for non-commercial use
PRO
Pro
/lean-ai-prd-team-pro
✅ Everything in Standard
⭐ Code Auditor (OWASP Top 10, Health Score /100)
⭐ BA Functional Point sizing (IFPUG Simplified)
⭐ Lean Waste Diagnosis [精益诊断] scenario
⭐ Value Assessor (ROI validation, Score /100)
IDE / Tool How to Use
Claude Code Native /lean-ai-prd-team slash command — one-line install
Cursor Paste SKILL.md → .cursorrules
Windsurf Paste SKILL.md → AGENTS.md
GitHub Copilot Paste → .github/copilot-instructions.md
JetBrains AI Settings → AI → Prompts → new prompt → paste SKILL.md
通义灵码 自定义指令 → 新建 → 粘贴 SKILL.md 全文
CodeBuddy 指令库 → 新建 → 粘贴 SKILL.md 全文
百度 Comate 设置 → 系统提示词 → 粘贴 SKILL.md 全文
Augment Code Instructions → Add Workspace Instructions → paste SKILL.md
Continue.dev ~/.continue/config.json → systemMessage → paste SKILL.md
Dify / Coze / FastGPT 系统提示词 → 粘贴 SKILL.md 全文
Pure API system role → SKILL.md content, user role → task with scenario prefix
Skill 1 · dev-team
8-Agent Scenario-to-MVP Delivery
You have a defined task — feature, refactor, or review. Scenario and value locked in first, code last. Token-efficient, minimum viable output.

已有明确任务，场景优先，最小MVP交付，节约Token。
git clone https://github.com/sky791016/lean-ai-dev-team \
  ~/.claude/skills/lean-ai-prd-team

# Then use:
/lean-ai-prd-team [全新项目] 你的任务
Skill 2 · lean-ai-agile-agent-team
Full Lifecycle Delivery System
Start from zero: goal → MVP → BA → architecture → sprint → ops → agent evolution.

从零开始，完整走完端到端交付闭环。
git clone https://github.com/sky791016/lean-ai-dev-team /tmp/lean-ai
cp -R /tmp/lean-ai/lean-ai-agile-agent-team \
  ~/.claude/skills/

# Bootstrap empty project:
bash .claude/skills/lean-ai-agile-agent-team/scripts/bootstrap-lean-ai-project.sh
Repository Structure
lean-ai-dev-team/
├── SKILL.md ← dev-team skill (10-agent rapid delivery)
├── README.md
├── references/ scenario-examples · ide-compatibility
└── lean-ai-agile-agent-team/ ← Full lifecycle skill (v2)
    ├── SKILL.md · README.md · MANIFEST.json
    ├── protocols/ lean-ai-delivery-protocol
    ├── roles/ 15 role definitions
    ├── scripts/ bootstrap · install · quality-check
    └── templates/ 30 document templates (00–30)
        input/ strategy/ discovery/ scenarios/ metrics/
        ba/ architecture/ delivery/ ops/ tasks/

Claude Code — dev-team (Rapid Delivery)

  1. Clone the skill

    # Clone into your Claude skills directory
    git clone https://github.com/sky791016/lean-ai-dev-team \
      ~/.claude/skills/lean-ai-prd-team

  2. Verify installation

    Open Claude Code in your project. The skill auto-loads from ~/.claude/skills/. No configuration needed.

  3. Invoke the team

    # Greenfield / Refactor / Review:
    /lean-ai-prd-team [全新项目] 你的任务描述

    /lean-ai-prd-team [重构优化] 你的任务描述

    /lean-ai-prd-team [项目评审] 你的任务描述

Claude Code — lean-ai-agile-agent-team (Full Lifecycle)

  1. Clone & install the skill

    git clone https://github.com/sky791016/lean-ai-dev-team /tmp/lean-ai
    cp -R /tmp/lean-ai/lean-ai-agile-agent-team ~/.claude/skills/

  2. Bootstrap your empty project

    # Run inside your project root:
    bash .claude/skills/lean-ai-agile-agent-team/scripts/bootstrap-lean-ai-project.sh

    # Add your brief:
    echo "Your project brief here" > input/brief.md

  3. Run the delivery pipeline

    # Goal alignment:
    Use lean-ai-agile-agent-team. Run goal alignment based on input/brief.md.

    # Discovery → Scenarios → MVP:
    Use lean-ai-agile-agent-team. Diverge requirements, converge, create scenario canvas, define MVP.

    # BA + Architecture + Sprint:
    Use lean-ai-agile-agent-team. Run BA pipeline, design architecture, start Sprint 1.

Other IDEs — Copy & Paste

Copy the contents of SKILL.md into your IDE's system prompt / instruction file. Then use the scenario prefixes in your chat.

# dev-team SKILL.md:
curl https://raw.githubusercontent.com/sky791016/lean-ai-dev-team/main/SKILL.md

# lean-ai-agile-agent-team SKILL.md:
curl https://raw.githubusercontent.com/sky791016/lean-ai-dev-team/main/lean-ai-agile-agent-team/SKILL.md

Lean AI Method · 精益AI方法论

The Methodology Behind the Team

Authored by Kai Shi (史凯), Founder of Lean AI Method. The complete framework for enterprise AI-native transformation.

📘

精益AI方法论

Lean AI Methodology

Enterprise AI-Native Transformation Framework

史凯 · Kai Shi

Founder of Lean AI Method

sky.kugua@gmail.com

📩 Request Full Whitepaper

Free for non-commercial use · Commercial licensing available

🎯

一心：以场景为核心

AI creates value only through specific business scenarios — not as abstract capability. Every deployment starts with scenario definition.

🔄

四闭环：持续进化机制

Value Loop, Data Loop, Model Loop, Ops Loop — four self-reinforcing cycles that make AI systems improve automatically over time.

🏗️

五工程：端到端实施

场景工程 → 数据工程 → 知识工程 → 智能体工程 → 运营工程. The complete engineering chain from concept to operations.

📊

AI成熟度五级模型

L0 No System → L1 Tool Trial → L2 Scenario POC → L3 Process Reengineering → L4 Platform → L5 AI-Native Operations.

Whitepaper Contents

✓ 企业AI转型三阶段判断框架
✓ 场景优先级评分矩阵（7维度加权）
✓ 精益AI场景卡标准模板
✓ ROI测算模型与案例（招采/客服/HR/研发）
✓ Clean Core + Cognitive Sidecar 架构图解
✓ AI运营指标体系（4类28项KPI）
✓ 企业AI转型0-36个月路线图

Author · 作者

Kai Shi (史凯) — Published Works

The Lean AI PRD Team skill is built on a decade of methodology research, published by the world's leading technical and academic publishers.

精益数据方法论

精益数据方法论

Lean Data Methodology

从业务场景出发的数据治理与价值实现方法论。已出版，广泛用于企业数据转型实践。

已出版 史凯 著
数据要素价值化蓝图

数据要素价值化蓝图

Data Factor Monetization Blueprint

数据要素如何从资源走向资产、实现商业价值的系统性方法与路线图。

已出版 史凯 著
📙
机械工业出版社

高质量数据建设指南

High-Quality Dataset Construction Guide

从场景到样本，企业高质量AI训练数据集建设的全流程工程方法与实践案例。

即将出版 史凯 著
SPRINGER
📕
Springer Nature

Lean AI Methodology
精益AI方法论

Enterprise AI-Native Transformation

The complete academic-grade methodology for enterprise AI transformation. International English edition — the foundation of this skill.

即将出版 Kai Shi
🎓

史凯 · Kai Shi — Founder of Lean AI Method

4本出版物 · 2家顶级出版社（机械工业出版社 + Springer Nature） · 精益数据方法论 → 精益AI方法论 系统方法论体系建设者 · 企业AI转型顾问 · 高质量数据集建设专家

联系作者 →

Pricing

Free to Use · Pro to Scale

Standard is free forever. Pro adds Code Audit, FP Sizing, and Value Assessment for teams that need to justify AI investment.

FREE

Standard

¥0

Free forever · Open source · Apache 2.0

  • ✓8 specialized AI agents
  • ✓3 scenarios: Greenfield / Refactor / Review
  • ✓Business Planner → PM → BA → Architect
  • ✓Frontend + Backend + Data parallel build
  • ✓4-loop Compliance PM sign-off
  • ✓Works with Claude Code, Cursor, Windsurf and 10+ IDEs
  • —Code Auditor (Pro only)
  • —BA FP Sizing / Value Assessor (Pro only)
Install Free →
PRO

Pro

¥199 / $29

per year · Commercial License Key · Written authorization

  • ★Everything in Standard
  • ★Code Auditor — OWASP Top 10 · Architecture Health Score /100
  • ★BA FP Sizing — IFPUG ILF/EIF/EI/EO/EQ · UFP total
  • ★Value Assessor — 4-dimension benefit · Realized Value Score /100
  • ★10 agents including Phase 0 Auditor + Phase 5b Value Assessor
  • ★Commercial use · License Key authorization

Purchase via

🇨🇳
微信 / 支付宝
扫码支付
💳
Stripe
Credit / Debit Card
☕
Ko-fi / BMaC
Buy Me a Coffee
🅿️
PayPal
Global · Instant
🔑 License Key Flow
Pay → Email receipt to sky.kugua@gmail.com → Receive License Key within 24 hrs
Buy Pro License →

☕ Support the Work (Optional Donation)

If Standard has been useful, a small tip keeps the methodology growing. No obligation — Standard is and always will be free.

☕ Ko-fi ☕ Buy Me a Coffee
🇨🇳 微信 / 支付宝
🅿️ PayPal

📦 Install via npm / npx

One command installs the skill to ~/.claude/skills/. No git required.

# Standard (free)
npx lean-ai-prd-team

# Pro (requires License Key)
npx lean-ai-prd-team --pro

The npm package clones the skill to ~/.claude/skills/. Node.js 14+ and git required. View on npm →

🚀 PRODUCT HUNT

Support the Future of AI-Native Development

If Lean AI PRD Team has changed how you think about AI-assisted development, show your support — every upvote helps the methodology reach more teams worldwide.

Upvote on Product Hunt ↑ Vote

Link updates when we launch — coming soon · Get notified

Star on GitHub ✉️ Contact Kai Shi 📘 Get Whitepaper
Share Your Story

Help Lean AI Reach the World

Lean AI is the first AI delivery methodology created in China, designed for global teams. Share your experience, show us a screenshot, or tell us what you built — every story helps this methodology reach more developers around the world.