Solution Design Review
Use this skill before proposing or implementing a solution when the user asks for a plan, diagnosis, implementation approach, best solution, root-cause analysis, refactor strategy, or tradeoff discussion.
Use When
- The user asks "how would you implement this?"
- The user asks for the best solution, not just a minimal patch
- The user asks why a bug happens and how to fix it
- The user wants to compare implementation options
- The task involves UI consistency, interaction state, component boundaries, architecture, or refactoring
- A quick local patch may work but could hide a deeper ownership or design issue
Do Not Use
- When the user clearly asks for a direct small edit and no discussion is needed
- For purely mechanical file changes with no design decision
- For commit message generation
- For user-facing documentation writing
Workflow
Classify the problem
- Bug fix
- Feature implementation
- Refactor
- UI/design consistency
- Interaction or state ownership
- Architecture or module boundary
- Platform or runtime behavior
Inspect the relevant context first
- Read the relevant code, config, or existing patterns before proposing a solution.
- Distinguish symptoms from root cause.
- Do not infer from class names or surface behavior alone when rendered output, event flow, or ownership boundaries matter.
Separate quick fixes from durable solutions
- If a quick fix exists, label it as such.
- If a better long-term solution exists, explain why it is better.
- Do not present a workaround as the best solution.
Recommend one approach
- State the recommended solution clearly.
- Explain the tradeoffs and blast radius.
- Avoid overengineering small changes, but do not default to minimal edits when the user asks for the best solution.
State implementation impact
- Files or modules likely affected
- Behavior changes
- Compatibility or migration concerns
- Verification plan
When approved, implement the recommended approach
- Do not silently downgrade to a quicker patch.
- If implementation reveals a better approach, pause and explain the change before switching strategy.
Key Rules
- Root cause before fix.
- Ownership boundary before patch.
- Rendered behavior before class-name assumptions.
- Tradeoffs before recommendation.
- Best solution means durable, maintainable, and appropriate to the project, not necessarily the largest refactor.
- When the user explicitly asks for the best solution in a development-stage codebase, optimize for the target design even if that requires refactoring; do not bias the solution toward backward compatibility with existing code unless the user asks for compatibility.
Output
For small cases, answer with:
原因:
方案:
影响范围:
验证方式:
For larger cases, include:
问题类型:
根因判断:
可选方案:
推荐方案:
实施范围:
验证计划:
1---2name: solution-design-review3description: Use when the user asks for an implementation plan, bug-fix approach, best solution, root-cause analysis, refactor strategy, tradeoff discussion, or wants to discuss how to solve a problem before coding.4---56# Solution Design Review78Use this skill before proposing or implementing a solution when the user asks for a plan, diagnosis, implementation approach, best solution, root-cause analysis, refactor strategy, or tradeoff discussion.910## Use When1112- The user asks "how would you implement this?"13- The user asks for the best solution, not just a minimal patch14- The user asks why a bug happens and how to fix it15- The user wants to compare implementation options16- The task involves UI consistency, interaction state, component boundaries, architecture, or refactoring17- A quick local patch may work but could hide a deeper ownership or design issue1819## Do Not Use2021- When the user clearly asks for a direct small edit and no discussion is needed22- For purely mechanical file changes with no design decision23- For commit message generation24- For user-facing documentation writing2526## Workflow27281. **Classify the problem**29 - Bug fix30 - Feature implementation31 - Refactor32 - UI/design consistency33 - Interaction or state ownership34 - Architecture or module boundary35 - Platform or runtime behavior36372. **Inspect the relevant context first**38 - Read the relevant code, config, or existing patterns before proposing a solution.39 - Distinguish symptoms from root cause.40 - Do not infer from class names or surface behavior alone when rendered output, event flow, or ownership boundaries matter.41423. **Separate quick fixes from durable solutions**43 - If a quick fix exists, label it as such.44 - If a better long-term solution exists, explain why it is better.45 - Do not present a workaround as the best solution.46474. **Recommend one approach**48 - State the recommended solution clearly.49 - Explain the tradeoffs and blast radius.50 - Avoid overengineering small changes, but do not default to minimal edits when the user asks for the best solution.51525. **State implementation impact**53 - Files or modules likely affected54 - Behavior changes55 - Compatibility or migration concerns56 - Verification plan57586. **When approved, implement the recommended approach**59 - Do not silently downgrade to a quicker patch.60 - If implementation reveals a better approach, pause and explain the change before switching strategy.6162## Key Rules6364- Root cause before fix.65- Ownership boundary before patch.66- Rendered behavior before class-name assumptions.67- Tradeoffs before recommendation.68- Best solution means durable, maintainable, and appropriate to the project, not necessarily the largest refactor.69- When the user explicitly asks for the best solution in a development-stage codebase, optimize for the target design even if that requires refactoring; do not bias the solution toward backward compatibility with existing code unless the user asks for compatibility.7071## Output7273For small cases, answer with:7475```text76原因:77方案:78影响范围:79验证方式:80```8182For larger cases, include:8384```text85问题类型:86根因判断:87可选方案:88推荐方案:89实施范围:90验证计划:91```