solving-problems
AI agent applies a 5-phase systematic framework for tackling complex problems when conventional approaches fail. Use when stuck, blocked, or troubleshooting issues.
📖 中文导读:AI agent applies a 5-phase systematic framework for tackling complex problems when conventional approaches fail. Use when stuck, blocked, or troubleshooting issues. 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。
name: solving-problems description: AI agent applies a 5-phase systematic framework for tackling complex problems when conventional approaches fail. Use when stuck, blocked, or troubleshooting issues. category: methodology triggers: - stuck - blocked - problem solving - troubleshooting - can't figure out
Solving Problems
Purpose
Apply a systematic 5-phase framework for complex problems:
- Define problem boundaries clearly before investigating
- Generate multiple hypotheses before testing any
- Test hypotheses efficiently with time-boxing
- Fix root causes, not just symptoms
- Prevent recurrence with process improvements
Quick Start
- Define (5-10 min) - Clarify what IS and IS NOT the problem, set success criteria
- Hypothesize (10-15 min) - Generate 3+ hypotheses BEFORE testing any
- Test (time-boxed) - Test cheapest hypothesis first, mark confirmed/eliminated
- Solve - Fix root cause, verify fix, add regression test
- Prevent - Analyze why it happened, implement preventive measures
Features
| Feature | Description | Guide |
|---|---|---|
| Problem Boundary | Define what IS vs IS NOT the problem | List symptoms, affected scope, working parts |
| Hypothesis Generation | Create 3+ theories before testing | Include evidence for/against, test cost |
| Prioritized Testing | Test highest-value hypotheses first | Score = evidence x (1/cost) |
| Root Cause Fix | Address underlying cause, not symptoms | Verify fix prevents recurrence |
| Prevention Plan | Stop future occurrences | Code, process, monitoring changes |
| Escalation Criteria | Know when to ask for help | Time-boxed out, all hypotheses eliminated |
Common Patterns
# Define Phase
BOUNDARY DEFINITION
What IS the problem:
- Specific error: [error message]
- Steps to reproduce: [steps]
- When started: [date/commit]
What is NOT the problem:
- What still works: [list]
- Red herrings eliminated: [list]
Success criteria:
- What does "solved" look like?
# Hypothesize Phase
H1: [Most likely] - Evidence: ___ - Test cost: Low
H2: [Second likely] - Evidence: ___ - Test cost: Medium
H3: [Less likely] - Evidence: ___ - Test cost: High
RULE: Generate hypotheses BEFORE testing any.
# Test Phase
Test H1: [15 min box]
Action: ___
Result: CONFIRMED / ELIMINATED / INCONCLUSIVE
# Risk Assessment Matrix
| Hypothesis | Probability | Impact | Test Effort | Priority |
|------------|-------------|--------|-------------|----------|
| H1 | High | High | Low | 1st |
| H2 | Medium | High | Medium | 2nd |
Use Cases
- Debugging complex bugs when conventional approaches fail
- Troubleshooting production incidents with unknown root causes
- Analyzing intermittent failures or race conditions
- Investigating performance regressions
- Resolving integration issues between systems
Best Practices
| Do | Avoid |
|---|---|
| Generate multiple hypotheses before testing | Testing first idea that comes to mind |
| Test cheapest hypothesis first when evidence equal | Spending 4 hours on 1 hypothesis |
| Time-box each phase | Rabbit holes without progress |
| Document failed approaches (valuable data) | Discarding hypotheses without testing |
| Verify root cause, not just symptoms | Fixing symptoms without understanding cause |
| Escalate when time-boxed out | Hero-coding beyond time limits |
| Add regression tests for every fix | Skipping prevention phase |
Related Skills
See also these related skill documents for complementary techniques:
debugging-systematically- Four-phase debugging frameworktracing-root-causes- Deep investigation techniquesthinking-sequentially- Structure reasoning as numbered thoughtsverifying-before-completion- Ensure fix is complete