Reflective Solution Framework

A structured thinking framework for AI agents that promotes critical analysis and iterative problem-solving.

Overview

This is an OpenCode agent skill that guides AI assistants through a systematic reflection process when analyzing problems, reviewing code, or proposing solutions. The framework ensures thorough evaluation before jumping to conclusions.

Assess → Propose → Nurture → Iterate

When to Use

Activate this skill when the user:

• Asks to analyze or review code/implementation
• Asks "what are the problems with this approach?"
• Asks "how would you solve this?"
• Asks for evaluation of a workflow or solution
• Wants deeper analysis or self-critique

Core Process

1. Assess

• Analyze the current implementation or situation
• Identify potential issues (edge cases, race conditions, error handling)
• Understand the user's real needs and constraints

2. Propose

• Provide concrete, actionable solutions
• Explain the implementation approach
• Clarify applicable scenarios and limitations

3. Nurture (Critical)

Key questions to ask yourself:

1. Does the solution solve the root problem, or just patch symptoms?
2. What edge cases are not considered?
3. Is there a better design approach?
4. What scenarios is this suitable for, and what are the limitations?
5. Is there a simpler, safer alternative?

Dimensions to evaluate:

• Reliability (concurrency, failure recovery)
• Performance impact
• Security and data integrity
• Maintainability
• User experience

4. Iterate

• Adjust or propose new solutions based on scrutiny
• If a better approach is found, replace confidently
• If the original solution is optimal, explain why

Response Format

## Assess
<current situation analysis>

## Solution
<proposed solution(s)>

## Nurture
### Pros
- ...

### Cons / Edge Cases
- ...

### Better Approach
<if found>

## Conclusion
<final recommendation>

Anti-Patterns

1. Don't rush to answer - fully evaluate first
2. Don't only list pros - must point out cons and edge cases
3. Don't cling to the first solution - replace if scrutiny reveals issues
4. Don't avoid calling something a "patch" - be honest about limitations

Tips

• If the issue involves data security or irreversible operations, prioritize the safest approach
• If the issue is performance optimization, provide benchmark suggestions
• If the user doesn't specify scenarios, proactively ask or cover multiple scenarios

Installation

This is an OpenCode skill. Install via:

opencode skill install reflective-skill

Files

• SKILL.en.md - English version
• SKILL.zh.md - Chinese version

License

MIT License

Contribution

Feel free to submit issues and pull requests.