From claude-skills
Use this skill when the user wants to improve, review, or update a TRIZ skill — including the XML primer prompt, SKILL.md routing description, and n8n MCP Server entry. Triggers on "TRIZ prompt verbessern", "skill verbessern", "TRIZ review", "prompt review", "lass uns heute triz_xyz verbessern", or any TRIZ tool name mentioned with improvement intent.
How this skill is triggered — by the user, by Claude, or both
Slash command
/claude-skills:triz-prompt-reviewThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Guides a structured improvement session for TRIZ tools. Each session covers all three artifacts together: **XML primer + SKILL.md + n8n MCP Server entry**. One skill per session.
Guides a structured improvement session for TRIZ tools. Each session covers all three artifacts together: XML primer + SKILL.md + n8n MCP Server entry. One skill per session.
Every TRIZ tool has three interdependent artifacts that must be kept in sync:
| Artifact | Repo / Location | Purpose |
|---|---|---|
SKILL.md | triz-skills repo | LLM routing — tells Claude when to invoke the tool |
| XML primer | triz-prompt-engineering repo | Full prompt behavior and process definition |
| n8n node | TRIZ MCP Server workflow | description field + ai_tool connection |
User names the skill to improve, or choose from the priority list below.
Read existing SKILL.md + XML from /Users/robert/Downloads/TRIZ-Skills-new/[skill]/ or directly from the GitHub repos.
Work through the XML checklist and SKILL.md checklist. Report all findings before touching anything.
Present planned changes clearly. Wait for user approval before implementing.
Edit in staging → deploy to GitHub → update n8n. See deployment pipeline below.
Suggest a concrete semi-automatic test scenario the user can run immediately.
| Check | Rule |
|---|---|
| Title length | ≤ 50 characters |
| Description length | ≤ 300 characters |
| Starters count | Exactly 2–3 |
| Instruction type attributes | Only type="general" and type="alternative" — no custom types |
| Required tags present | <Goal>, <Role>, <Output_Format>, <User_Input> all exist |
| Correct tag name | <Output_Format> not <Output> |
| Example structure | Each <Example title="..."> has <Problem> + <Solution> children |
| Three-mode pattern | Mode selection as Step 1 in type="general"; details in type="alternative" |
| Output completeness | All required output elements explicitly stated in <Output_Format> |
In Instruction type="general", Step 1 must read:
Ask the user which working mode they prefer: Automatic (generate immediately using own assumptions, state them), Semi-automatic (ask [N] targeted questions, then generate), or Interactive (step by step with user confirmation at each stage).
In Instruction type="alternative":
| Check | Rule |
|---|---|
name field | Present in frontmatter |
description field | Contains specific trigger phrases users would actually say |
| Trigger keywords | Precise, non-overlapping with other TRIZ skills |
| Negative abgrenzung | Add "Does NOT handle X" when another skill covers similar territory |
| Sync with n8n | SKILL.md description = n8n node description (must be identical) |
| Check | Rule |
|---|---|
| Node description | Matches SKILL.md description exactly |
| Connection type | ai_tool — never main |
| Fix syntax if wrong | removeConnection then addConnection with sourceOutput: "ai_tool" |
| Workflow ID | QDDXjTgvUQOpJqou (TRIZ MCP Server) |
Staging: /Users/robert/Downloads/TRIZ-Skills-new/[skill]/
GitHub 1: /Users/robert/Documents/GitHub/triz-prompt-engineering/
prompts/technical_triz/[skill]/[skill].xml
⚠ Branch protection on main → create branch + PR via gh
GitHub 2: /Users/robert/Documents/GitHub/triz-skills/[skill]/SKILL.md
✓ Direct push to main OK
n8n: Workflow QDDXjTgvUQOpJqou (TRIZ MCP Server)
→ Update description on tool node (must match SKILL.md)
→ Verify ai_tool connection
Skills with clear process logic that benefit most from the three-mode pattern:
triz_contradiction_solver — Widerspruchsmatrix + ARIZ, klare Schrittfolgetriz_function_analysis — mehrstufiger Prozess, sehr häufig genutzttriz_cause_effect_chain_analysis — iterative Analysetriz_trimming / triz_interactive_trimming — Dopplung prüfen, evtl. zusammenführentriz_resource_analysistriz_idealitytriz_76_standard_solutions| Fact | Value |
|---|---|
| TRIZ MCP Server workflow ID | QDDXjTgvUQOpJqou |
| triz-prompt-engineering | Branch protection on main → always PR |
| triz-skills | Direct push to main OK |
| Staging area | /Users/robert/Downloads/TRIZ-Skills-new/ |
| n8n ai_tool connection | addConnection needs sourceOutput: "ai_tool" |
| triz-prompt-engineering path | prompts/technical_triz/[skill]/[skill].xml |
| triz-skills path | [skill]/SKILL.md |
npx claudepluginhub robert-adunka/claude-skills --plugin claude-skillsGuides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.