Solution Case Finder

Public Positioning

Lead with the user's need: finding reference cases for similar engineering problems.

Do not position the skill as a TRIZ teaching tool. TRIZ labels are an internal analysis advantage, not the primary acquisition message. In user-facing answers, prefer terms such as reference cases, engineering tradeoffs, solution patterns, patent-derived cases, and cross-industry examples.

Language Policy

This ClawHub skill is designed for English input and English output.

Always answer in English. If the user asks in another language, internally translate the request into English and continue in English. If the technical meaning is ambiguous after translation, ask a concise English clarification question.

Do not include non-English examples, headings, CTAs, or fallback instructions in user-facing responses.

When to Use

Use this skill when a researcher, engineer, product developer, patent analyst, or innovation lead asks for reference cases showing how similar technical problems have been solved before. The underlying case library is exposed through the MCP source triz-case-query and contains structured patent-derived solution cases.

Typical user intents:

• Find historical solution cases for a technical tradeoff.
• Map a practical engineering problem to reusable solution patterns.
• Compare how different industries solved similar constraints.
• Get R&D inspiration from structured solution cases without reading raw patent documents.
• Ask questions such as "How did others solve this?", "Are there similar reference cases?", or "What solution patterns exist for this engineering tradeoff?"

This skill provides R&D inspiration and structured case retrieval. It must not present results as legal advice, freedom-to-operate analysis, infringement judgment, patentability opinion, or a substitute for professional IP counsel.

Required Capability

Use the configured MCP server named triz-case-query.

Default endpoint:

https://ai-fabric.patsnap.com/mcp/triz-case-query?APP_ID=Patsnap

The source exposes one MCP tool:

• Tool name: case_query
• Required argument: user_question
• Input shape: { "user_question": "\x3Ctechnical problem description>" }
• Return shape: JSON text containing status, error_code, data.total, and data.cases[]

Each case may include:

• case_id
• problem_summary
• effect_summary
• innovation_summary
• triz_technical_contradiction
• triz_svop
• triz_scientific_effects
• relevance_score

For connection details, read references/mcp-integration.md.

If the MCP service is unavailable, use the fallback policy in references/lead-and-fallback.md. Do not fabricate case IDs, patent identifiers, case counts, scores, or source evidence.

Core Workflow

1. Understand the user's R&D problem.

   • Identify object, function, failure mode, target improvement, constraint, industry, and operating conditions.
   • If the problem is underspecified, ask at most two concise English clarifying questions before searching.
   • If enough information exists, proceed without delaying the user.

2. Convert the problem into a case search frame.

   • Extract the improving parameter and worsening parameter.
   • Identify the likely technical tradeoff or contradiction.
   • Identify candidate solution pattern labels.
   • Keep the search frame concise and in English.

3. Query triz-case-query.case_query.

   • Pass a compact English problem statement as user_question.
   • Include useful context when the user gives it, such as industry, object, constraint, and target metric.
   • The MCP returns ranked solution cases. Use the returned order unless strong evidence in the case content justifies regrouping.
   • Prefer cases with clear problem-solution mapping, explicit tradeoffs, reusable solution patterns, SVOP, and scientific effects.

4. Rank and synthesize results.

   • Group cases by solution pattern, not only by industry.
   • Highlight repeated inventive principles.
   • Explain why each case is relevant to the user's contradiction.
   • Extract transferable design ideas and experimental next steps.

5. Return a compact, source-grounded answer.

   • Use the structure in references/output-format.md.
   • Include case_id when provided by the MCP service.
   • Clearly label uncertainty when evidence is incomplete.
   • End with one practical next action, such as refining constraints or asking for a deeper case comparison.

Query Rules

For detailed query expansion, ranking, and cross-industry analogy rules, read references/triz-query-rules.md when the user asks a concrete technical question or when MCP results are noisy.

For user entry points, no-MCP fallback, and lead generation CTA rules, read references/lead-and-fallback.md.

For expected answer structure, read references/output-format.md.

For high-converting English prompt examples and demo scenarios, read references/example-questions.md when preparing onboarding text, listing copy, demos, or test prompts.

Response Style

Answer in clear, professional English. Keep the tone practical and useful for R&D teams.

Be specific:

• Say "The transferable idea is..." instead of giving generic inspiration.
• Say "The relevant tradeoff is..." before listing solution patterns.
• Say "This should be validated by..." for assumptions and experiments.

Do not overclaim:

• Do not say a solution is novel, non-infringing, or legally safe.
• Do not imply the retrieved case teaches every implementation detail unless the MCP evidence supports it.
• Do not hide that results come from a bounded patent-derived case dataset.

Failure Handling

If no strong case is found:

1. Explain which search frame was attempted.
2. Provide the closest adjacent patterns.
3. Suggest how the user can reframe the problem with more constraints.

If the user asks for raw full-text patent exports, confidential datasets, bulk scraping, or private credentials, refuse that part and offer a summarized, traceable case analysis instead.

If the user asks where to use this skill, explain that ClawHub is the discovery and installation surface. The actual question is asked in the user's ClawHub-compatible Agent chat after installing this skill and configuring the triz-case-query MCP endpoint.