一项Claude / AI代理技能，用于解决Altshuller的TRIZ（发明性问题解决理论）中的工程问题。用基于专利衍生模式的算法问题解决取代了基于折衷的头脑风暴。

它的作用

在工程矛盾情况下——*“改进X会让Y变差”或“这部分必须既是A又非A**——该技能会产生3–5个具体的创意概念，这些概念有明确的理想度评分，而非妥协方案。它使用：

• 理想最终结果（IFR） 框架
• Altshuller从~20万项专利中提炼出的39个工程参数和40个发明原理
• 完整的39×39矛盾矩阵（1190个填充单元格，锚定验证对照Altshuller 1985）
• Su-场分析，采用标准76标准解（5类分类法，Su-域代数符号）
• 物理矛盾的分离原则（空间/时间/条件/系统层面）
• ARIZ-85C 快速通过未达理想度条时的深度程序
• 路线图问题的工程系统演进趋势**

何时使用

triz-engineering-solver

A Claude / AI-agent skill for solving engineering problems with Altshuller’s TRIZ (Theory of Inventive Problem Solving). Replaces compromise-driven brainstorming with algorithmic problem solving over a corpus of patent-derived patterns.

What it does

Given an engineering contradiction — “improving X makes Y worse” or “this part must be both A and not-A” — the skill produces 3–5 concrete inventive concepts with explicit ideality scores rather than compromise solutions. It uses:

• Ideal Final Result (IFR) framing
• The 39 engineering parameters and 40 inventive principles distilled by Altshuller from ~200,000 patents
• The full 39×39 Contradiction Matrix (1190 populated cells, anchor-verified against Altshuller 1985)
• Su-Field analysis with the canonical 76 Standard Solutions (5-class taxonomy, Su-Field algebra notation)
• Separation principles (space / time / condition / system-level) for physical contradictions
• ARIZ-85C deep procedure when the quick pass fails the ideality bar
• The 8 trends of engineering system evolution for roadmapping questions

When to use

• Engineering trade-offs where improving parameter A degrades parameter B (technical contradiction)
• A single element must exhibit opposite properties (physical contradiction)
• Design bottleneck where conventional optimisation has plateaued
• System redesign aimed at the Ideal Final Result

When NOT to use

• Pure software architecture without a physical analogue
• UX / interaction design
• Business or organisational strategy
• Open brainstorming with no concrete contradiction identified

The skill refuses with reframe when invoked on out-of-scope problems — see examples/anti_example_misframed.md.

Install

For Claude Code:

git clone https://github.com/Antropocosmist/triz-engineering-solver \
  ~/.claude/skills/triz-engineering-solver

For other agent runtimes (Anthropic SDK, OpenAI Assistants, LangGraph, custom): point the agent’s system prompt or tool router at SKILL.md and grant filesystem read access to this directory.

Repo layout

.
├── README.md
├── LICENSE                          MIT
├── CONTRIBUTING.md                  anchor-cell verification protocol, style rules
├── SKILL.md                         entry point — workflow, triggers, output template
├── resources/                       lazily-loaded reference data
│   ├── 39_parameters.md
│   ├── 40_principles.md
│   ├── 76_standard_solutions.md     full 5-class taxonomy with Su-Field algebra
│   ├── contradiction_matrix.json    full 39×39 Altshuller matrix (1190 cells)
│   ├── separation_principles.md     decision procedure for physical contradictions
│   ├── ariz_85c.md                  9-part deep-analysis procedure
│   ├── evolution_trends.md          8 trends + S-curve framing
│   ├── glossary.md