sequential-thinking
Structured reasoning tools for complex problem analysis with observation, hypothesis, analysis, and conclusion steps. Use when analyzing complex problems, debugging difficult issues, or making important decisions.
When & Why to Use This Skill
The Sequential Thinking skill is a powerful reasoning framework designed for Claude to tackle complex problem analysis through a structured, multi-step process. By implementing a systematic workflow of observation, hypothesis formation, and evidence-based analysis, this skill ensures that the AI avoids jumping to conclusions and instead provides logical, well-documented solutions for technical troubleshooting and critical decision-making.
Use Cases
- Technical Root Cause Analysis: Systematically identifying the source of software bugs or system failures by documenting facts, forming hypotheses, and testing them against logs.
- Complex Decision Making: Evaluating multiple strategic options by using branching logic to explore alternative scenarios and their potential outcomes.
- Architecture and Logic Review: Breaking down intricate system designs or logical arguments into verifiable steps to ensure consistency and identify potential flaws.
- Incident Response: Providing a structured 'thought trace' during live system outages to ensure all variables are considered before applying a fix.
| name | sequential-thinking |
|---|---|
| description | Structured reasoning tools for complex problem analysis with observation, hypothesis, analysis, and conclusion steps. Use when analyzing complex problems, debugging difficult issues, or making important decisions. |
| allowed-tools | Read, Write |
Sequential Thinking Skill
Version: 1.0.0 Purpose: Structured reasoning for complex problem analysis
Triggers
| Trigger | Examples |
|---|---|
| Analyze | "analyze problem", "think through", "問題分析" |
| Debug | "debug this", "investigate issue", "調査" |
| Decide | "help me decide", "evaluate options", "判断支援" |
| Reason | "reason about", "think step by step", "段階的に考える" |
Integrated MCP Tools
| Tool | Purpose |
|---|---|
think_step |
Record a reasoning step |
think_branch |
Create alternative thinking branch |
think_summarize |
Summarize thinking session |
Step Types
| Type | Purpose | Example |
|---|---|---|
observation |
Gather facts | "The error occurs at startup" |
hypothesis |
Form theory | "Configuration may be incorrect" |
analysis |
Evaluate | "Testing hypothesis against logs" |
conclusion |
Final result | "Root cause identified as X" |
question |
Open question | "What is the expected behavior?" |
Workflow: Problem Analysis
Phase 1: Observation
Step 1.1: Gather Facts
Use think_step with:
- thought: "Observed: [specific observation]"
- type: "observation"
- confidence: 0.9 (high certainty)
Step 1.2: Document Context
Multiple think_step calls for each fact:
- System state
- Error messages
- Recent changes
- User reports
Phase 2: Hypothesis Formation
Step 2.1: Form Hypotheses
Use think_step with:
- thought: "Hypothesis: [potential cause]"
- type: "hypothesis"
- confidence: 0.5 (initial guess)
Step 2.2: Alternative Hypotheses
Use think_branch to explore:
- Different root causes
- Alternative explanations
- Edge cases
Phase 3: Analysis
Step 3.1: Test Hypotheses
Use think_step with:
- thought: "Testing: [method and result]"
- type: "analysis"
- confidence: [adjusted based on evidence]
Step 3.2: Eliminate Options
Use analysis steps to:
- Confirm or refute each hypothesis
- Document evidence
- Adjust confidence levels
Phase 4: Conclusion
Step 4.1: Draw Conclusions
Use think_step with:
- thought: "Conclusion: [final determination]"
- type: "conclusion"
- confidence: 0.9 (based on evidence)
Step 4.2: Summarize Session
Use think_summarize with:
- sessionId: Current session
- includeAlternatives: true
Branching Strategy
When to Branch
- Multiple valid hypotheses
- Need to explore alternatives
- Complex decision with trade-offs
Branch Usage
Use think_branch with:
- sessionId: Current session
- branchName: "alternative-approach"
- fromStep: Step number to branch from
Confidence Levels
| Level | Value | Meaning |
|---|---|---|
| Certain | 0.9-1.0 | Strong evidence, verified |
| Likely | 0.7-0.9 | Good evidence, probable |
| Possible | 0.5-0.7 | Some evidence, uncertain |
| Unlikely | 0.3-0.5 | Weak evidence |
| Doubtful | 0.0-0.3 | Little to no evidence |
Example Session
1. [observation] Error: "Connection refused" on port 5432
2. [observation] PostgreSQL service status: stopped
3. [hypothesis] Database service crashed
4. [hypothesis] Port conflict with another service
5. [analysis] Checking service logs... no crash, clean shutdown
6. [analysis] Checking port usage... no conflict
7. [conclusion] Service was manually stopped, needs restart
Best Practices
✅ GOOD:
- Start with observations
- Form multiple hypotheses
- Document evidence for/against
- Update confidence as you learn
❌ BAD:
- Jump to conclusions
- Ignore contradicting evidence
- Single hypothesis bias
- Skip documentation
Checklist
- Problem clearly stated
- Facts gathered (observations)
- Multiple hypotheses formed
- Evidence documented
- Alternatives explored (branches)
- Conclusion supported by evidence