triz-skills

Universal TRIZ Skills for Claude – open, structured, and future-proof

This repository contains a growing, curated library of TRIZ (Theory of Inventive Problem Solving) Skills for Anthropic’s Claude – covering both technical and business TRIZ, from Contradiction Analysis and Function Analysis to Trimming, the 76 Standard Solutions, Ideality, and Resource Analysis.

Every Skill is kept unzipped, one folder per Skill, so it can be read, edited, and extended with any standard text editor (e.g. VS Code, Notepad++, BBEdit) and collaborated on with normal Git workflows.

Maintained by Truinorva. Companion repository to triz-prompt-engineering.

---

🌐 Purpose

Our goal is to make TRIZ knowledge usable and sustainable in the age of AI – platform-independent, transparent, and free from proprietary constraints.

This approach allows for:

• Reusable, agent-ready TRIZ methods that Claude can activate on demand
• Future-proof archiving and documentation in plain text
• Simplified integration into company-specific innovation workflows

---

🔧 Key Features

• ✅ Self-contained Skills – each Skill is one folder with a SKILL.md plus optional references/ and assets/
• 📦 ZIP = release – develop unzipped in Git, ship as ZIPs built on demand and attached to a GitHub Release; no archives are stored in the repo (see Releasing)
• 🧭 Two families – technical-triz/ for engineering, business-triz/ for organizational and people-centric problems
• 📖 English-first, term-bilingual – content is written in English, with key terms also given in another language (e.g. German) so the AI handles terminology correctly
• ⚖️ MIT License – open use, including commercial applications
• 📁 Git-based collaboration – with versioning, issue tracking, and a curated main branch
• 🔀 Pull-request workflow – changes are made in a feature branch and merged into main via PR

---

🧩 What is a Claude Skill?

A Skill is a folder that packages instructions, reference data, and optional helper scripts so Claude can reliably perform a specialized task. At minimum it contains a SKILL.md file with YAML frontmatter (a name and a description). Claude reads the description to decide when to activate the Skill, then follows the body of SKILL.md to do the work, loading files from references/ only as needed.

Claude distributes and installs Skills as ZIP archives. This repository keeps them unzipped for development and collaboration – releasing a Skill is simply zipping its folder.

---

📂 Directory Structure

triz-skills/
│
├── docs/                         # placeholder; the authoring & packaging guidance is in this README
├── scripts/
│   └── build_skills.py           # packages every Skill folder into a release ZIP
├── .github/workflows/
│   └── release.yml               # validates on every push/PR, publishes ZIPs on a tag
│
├── business-triz/                # TRIZ for organizational, service & people-centric problems
│   └── /
│       ├── SKILL.md
│       ├── references/           # (optional) knowledge Claude reads on demand
│       └── assets/               # (optional) files the Skill ships and hands on
│
├── technical-triz/               # classical engineering-oriented TRIZ
│   └── /
│       ├── SKILL.md
│       ├── references/           # (optional) knowledge Claude reads on demand
│       └── assets/               # (optional) files the Skill ships and hands on
│
├── LICENSE
└── README.md

dist/                             # build output — git-ignored, never committed

This layout mirrors triz-prompt-engineering: the same separation of business vs. technical TRIZ, with each tool isolated in its own self-contained folder, and a shared docs/ folder for guides.

---

📚 Skill Index

Business TRIZ (business-triz/)

Technical TRIZ (technical-triz/)

---

▶️ Using a Skill with Claude

1. Download the Skill’s ZIP from the latest release — one archive per Skill, ready to install.
2. Install it into your Claude environment (e.g. Claude, Cowork, Claude Code) as a Skill.
3. Describe your problem; Claude activates the matching Skill based on its description.

Install one Skill per ZIP — Claude installs a Skill from an archive rooted in that Skill’s own folder, so there is deliberately no combined “all Skills” archive to import.

During development you can also point Claude directly at an unzipped Skill folder in this repo.

---

📦 Releasing (packaging the Skills)

ZIPs are build artefacts, not repository content. They are never committed — dist/ and *.zip are git-ignored. Every archive is rebuilt from the Skill folders and published as a GitHub Release asset, so each Skill exists exactly once in Git: as its editable folder.

Archive layout

Claude installs a Skill from an archive whose root entry is the Skill folder itself:

contradiction-solver.zip
└── contradiction-solver/          # folder name == frontmatter `name`
    ├── SKILL.md
    ├── references/...
    └── assets/...                 # if the Skill ships any

One archive holds exactly one Skill. That single top-level folder is the format, so an archive bundling several Skills side by side cannot be imported into Claude at all. An all-in-one triz-skills-all.zip was briefly offered with v1.0.0 and withdrawn for exactly this reason — the build no longer produces one, and no release carries one. Please do not reintroduce it. Installing the full library means installing the individual ZIPs.

Cutting a release

Tag the commit and push the tag — the Package Skills workflow builds all archives and attaches them to the release:

git tag -a v1.0.0 -m "Release v1.0.0" && git push origin v1.0.0

The release then carries one ZIP per Skill plus SHA256SUMS.txt — and deliberately no combined “all Skills” archive, since Claude cannot import one (see Archive layout).

Building locally

python scripts/build_skills.py

This writes the same archives into dist/. Useful flags:

The build packages only Git-tracked files, so untracked scratch files and OS junk (.DS_Store, Thumbs.db) can never leak into a release.

Packaged content is always byte-identical to the committed bytes, on every platform: entries are sorted, carry a fixed timestamp, and text files are normalized to LF, so a Windows checkout (which Git hands you with CRLF) produces the same file contents as a Linux one. The compressed archive itself can still differ in size between machines, because that depends on the zlib build behind Python — rebuilding in the same environment is byte-identical.

What the build enforces

The same validation runs locally and in CI (on every push and pull request), so a Skill that cannot be packaged fails before it reaches a release:

• SKILL.md exists and starts with YAML frontmatter carrying name and description
• name is lowercase-hyphenated and matches the folder name
• no two Skills share a name
• every Skill has committed files
• each finished archive really contains /SKILL.md at exactly one top-level folder

---

✍️ Authoring a Skill — structure & tips

Required: SKILL.md with YAML frontmatter.

---
name: contradiction-solver
description: "TRIZ Contradiction Solver and 40 Inventive Principles — resolves
  engineering and physical contradictions ... Use this skill whenever the user
  mentions 'contradiction', 'inventive principles', '40 principles', ..."
---

# Human-readable title

Body: step-by-step instructions Claude should follow, what to ask the user,
which reference files to read, and how to present results.

Things to observe:

• name — lowercase, hyphenated, unique, and matching the folder name. This is the Skill’s identifier.
• description is the trigger. It is the only thing Claude sees when deciding whether to activate a Skill. Write it in the third person, state what the Skill does, and list the words/phrases that should activate it (“Use this skill when the user mentions …”). Be specific to avoid both false triggers and missed triggers. Aim for roughly 1–3 sentences.
• Keep SKILL.md lean; push detail into references/. Claude loads SKILL.md up front but reads reference files only when needed. Put long tables, matrices, examples, and images (e.g. the 40 Principles, the Contradiction Matrix, the 76 Standard Solutions) in references/ and tell SKILL.md to read them on demand. This keeps activation cheap and the instructions focused.
• references/ for knowledge, assets/ for files that are handed on. references/ holds what Claude reads to do the work — tables, matrices, method descriptions, examples. assets/ holds what the Skill ships and passes to the user: templates to fill in, source PDFs, binaries. See innovation-checklist, whose ISQ questionnaires and Excel templates live in assets/. Both are optional, and a Skill may have either, both, or neither.
• Reference paths are relative to the Skill folder (e.g. references/40_Inventive_Principles_EN.md, assets/TRU_ISQ_EN.xlsx). Never use absolute or machine-specific paths.
• Bundle data with the Skill. Anything the Skill needs (CSVs, Markdown tables, PNG diagrams, helper scripts) lives inside the folder so the ZIP is self-contained.
• Write in English first. English is the primary language for all SKILL.md instructions and documentation.
• Add key terms in another language to anchor terminology. Where a precise term or phrasing matters, include the equivalent in another language (e.g. German) inline so the AI maps and handles it correctly. See contradiction-solver for examples — the contradiction formulation gives both IF … THEN … BUT … and WENN … DANN … ABER …, and names both Altshuller Matrix and Altschuller Matrix. Parallel reference files (e.g. *_EN.md / *_DE.md) are welcome for larger bilingual data sets.
• One Skill = one folder = one job. Prefer several focused Skills over one giant Skill; it improves trigger accuracy and reuse.
• License header. Add a short license/attribution comment near the top of SKILL.md referencing the repo-root LICENSE.

A fuller authoring guide and a Skill template will live in docs/.

---

🤝 Collaboration Welcome

This project is open for contributions – while keeping the main branch curated to ensure quality, transparency, and traceability.

Whether you’re an AI developer, TRIZ expert, educator, or innovation practitioner – feel free to fork, improve, and suggest new Skills. Let’s build a shared, evolving toolkit for AI-enhanced TRIZ applications.

• 💡 Before you start: read Authoring a Skill and Releasing in this README. docs/ is still a placeholder — the fuller guides announced there do not exist yet.
• 🧩 Adding a Skill: create a new folder under business-triz/ or technical-triz/, following the structure above. Make sure SKILL.md has a clear, well-triggered description and that all reference paths are relative.
• 🚩 Contribution rule: make commits on a branch other than main (e.g. development) and submit them via a pull request before they are merged into main.

---

⚖️ License

Released under the MIT License — see LICENSE. © 2026 Truinorva.