forge-skill
peterfei/forge-skill
锻造任何领域的思维工具 — 不是角色扮演一个智者,而是锻造一把可执行的思维工具。基于 Agent Skills 开放协议,可在 Claude Code、Cursor 等 50+ runtime 中运行。
Instalação
npx convez-ai peterfei/forge-skill1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
README
A tradução para o português deste README ainda está na fila.
Forge Skill — 锻造任何领域的思维工具
不是角色扮演一个智者,而是锻造一把可执行的思维工具。 输入方法论名称或模糊需求,自动深度调研 → 方法论结构化提炼 → 生成可运行的思维工具 Skill。
简体中文
这是什么?
Forge Skill 是一个基于 Agent Skills 开放协议的思维工具锻造引擎。它将经典方法论(第一性原理、系统思维、TRIZ 等)转化为可执行的 AI Skill,让 AI 不再"扮演专家",而是按照方法论的精确流程帮你分析问题。
一句话:把方法论变成你随时调用的认知瑞士军刀。
核心特性
- 锻造引擎:输入方法论名 → 6 路并行调研 → 三重验证提炼 → 生成可执行 Skill
- 诊断推荐:不确定用什么方法?描述你的困惑,引擎推荐最匹配的方法论
- 误用检测:每个 Skill 内置误用检测器,主动警告方法论被错误应用
- 诚实边界:明确标注方法的局限,不伪装万能
- 跨 Runtime:兼容 Claude Code、Codex CLI、Cursor、Gemini CLI 等 50+ 运行时
- 多语言:中文、英语、日语、韩语等
快速开始
# 安装(需要 Agent Skills 运行时,如 Claude Code)
npx skills add peterfei/forge-skill
# 直接锻造(明确方法论)
> 锻造一个第一性原理
# 诊断推荐(模糊需求)
> 我的产品留存率持续下降,不知道怎么办
锻造流程
Phase 0 入口分流 → 直接锻造 or 诊断推荐
Phase 1 多源采集 → 6 路并行调研(文献/案例/批评/跨域/工具/时间线)
Phase 1.5 调研 Review → 展示摘要,确认质量
Phase 2 方法论提炼 → 三重验证核心原理 + 操作流程 + 适用边界
Phase 2.5 提炼确认 → 展示摘要,确认方向
Phase 3 Skill 构建 → Agentic Protocol + 误用检测器 + 诚实边界
Phase 4 质量验证 → 已知案例 + 边缘案例 + 误用检测验证
Phase 5 双 Agent 精炼 → 方法论评审 + 可用性评审
预锻造方法论库
| # | 方法论 | 领域 | 一句话定位 |
|---|---|---|---|
| F1 | 第一性原理 | 解决问题/创新 | 回归基本事实,从物理极限重新推导 |
| F2 | 系统思维 | 复杂系统分析 | 用因果回路图看清整体结构和反馈循环 |
| F3 | 增长黑客 | 产品增长/留存 | 以数据驱动的实验循环寻找增长杠杆 |
| F4 | 精益创业 | 创业验证 | 用最小可行产品快速验证假设 |
| F5 | 设计思维 | 产品设计/创新 | 以用户共情为起点的五步创新框架 |
| F6 | TRIZ | 工程创新/矛盾解决 | 40 个发明原理系统化解决技术矛盾 |
| F7 | 费曼学习法 | 学习/教学 | 用教别人的方式检验自己是否真正理解 |
| F8 | 改善(Kaizen) | 持续改进/管理 | 小步快跑的渐进式优化哲学 |
| F9 | 事前验尸(Premortem) | 风险预判/决策 | 假设项目已失败,反推所有可能死因 |
| F10 | 五力分析 | 竞争战略 | 从五个维度评估行业竞争格局 |
| F11 | 双钻模型 | 设计流程/问题定义 | 发散→收敛→发散→收敛的结构化流程 |
| F12 | 奥卡姆剃刀 | 决策/简化 | 在同等解释力下选择最少假设的方案 |
生态
Forge Skill 是三项目生态的一部分:
nuwa-skill(女娲) → 蒸馏人的认知框架(WHO)
forge-skill(熔炉) → 锻造方法论工具(HOW)
darwin-skill(达尔文)→ 优化已有 Skill(BETTER)
- nuwa-skill — 蒸馏名人思维方式
- darwin-skill — 持续进化优化 Skill
License
English
What is this?
Forge Skill is a thinking tool forge engine based on the Agent Skills open protocol. It transforms classic methodologies (First Principles, Systems Thinking, TRIZ, etc.) into executable AI Skills — making AI follow precise methodological processes instead of "role-playing experts."
In one sentence: Turn methodologies into cognitive Swiss Army knives you can invoke anytime.
Core Features
- Forge Engine: Input a methodology name → 6-way parallel research → triple-verified extraction → generate executable Skill
- Diagnostic Recommendation: Not sure which method to use? Describe your problem, the engine recommends the best match
- Misuse Detection: Each Skill has a built-in misuse detector that proactively warns when a methodology is misapplied
- Honest Boundaries: Clearly labels method limitations — never pretends to be universal
- Cross-Runtime: Compatible with Claude Code, Codex CLI, Cursor, Gemini CLI, and 50+ runtimes
- Multilingual: Chinese, English, Japanese, Korean, and more
Quick Start
# Install (requires an Agent Skills runtime, e.g., Claude Code)
npx skills add peterfei/forge-skill
# Direct forge (specific methodology)
> Forge a First Principles skill
# Diagnostic recommendation (vague need)
> My product retention keeps dropping, I don't know what to do
Forge Pipeline
Phase 0 Entry Routing → Direct forge or diagnostic recommendation
Phase 1 Multi-source Collection → 6-way parallel research (literature/cases/criticism/cross-domain/tools/timeline)
Phase 1.5 Research Review → Show summary, confirm quality
Phase 2 Methodology Extraction → Triple-verified core principles + operational flow + applicability boundaries
Phase 2.5 Extraction Confirmation → Show summary, confirm direction
Phase 3 Skill Construction → Agentic Protocol + misuse detector + honest boundaries
Phase 4 Quality Validation → Known cases + edge cases + misuse detection verification
Phase 5 Dual Agent Refinement → Methodology review + usability review
Pre-forged Methodology Library
| # | Methodology | Domain | One-line Description |
|---|---|---|---|
| F1 | First Principles | Problem-solving/Innovation | Reduce to fundamental truths, reason up from physical limits |
| F2 | Systems Thinking | Complex Systems | See overall structure and feedback loops through causal loop diagrams |
| F3 | Growth Hacking | Product Growth/Retention | Data-driven experimentation cycles to find growth levers |
| F4 | Lean Startup | Startup Validation | Validate hypotheses rapidly with minimum viable products |
| F5 | Design Thinking | Product Design/Innovation | Five-step innovation framework starting from user empathy |
| F6 | TRIZ | Engineering/Contradiction Resolution | 40 invention principles for systematic resolution of technical contradictions |
| F7 | Feynman Learning Technique | Learning/Teaching | Test your understanding by teaching it to others |
| F8 | Kaizen | Continuous Improvement | Small-step incremental optimization philosophy |
| F9 | Premortem | Risk Anticipation/Decision | Assume the project failed, reverse-engineer all possible causes of death |
| F10 | Five Forces | Competitive Strategy | Assess industry competitive landscape from five dimensions |
| F11 | Double Diamond | Design Process/Problem Definition | Diverge→converge→diverge→converge structured process |
| F12 | Occam's Razor | Decision/Simplification | Choose the solution with fewest assumptions given equal explanatory power |
Ecosystem
Forge Skill is part of a three-project ecosystem:
nuwa-skill → Distill human cognitive frameworks (WHO)
forge-skill → Forge methodology tools (HOW)
darwin-skill → Optimize existing Skills (BETTER)
- nuwa-skill — Distill notable thinkers' mindsets
- darwin-skill — Continuously evolve and optimize Skills