sequential-thinking（顺序思维）

Sequential Thinking Skill

Installation

Optional standalone executor: The skill can run .claude/tools/optimization/sequential-thinking/executor.py. Requirements:

• Python 3.10+: python.org or winget install Python.Python.3.12 (Windows), brew install python@3.12 (macOS).
• MCP Python package: pip install mcp
• If the executor is not present or MCP is not installed, the skill still provides in-context guidance and exits successfully.

Full functionality: The Python script .claude/tools/optimization/sequential-thinking/executor.py must exist for standalone executor runs. Run node.claude/tools/cli/doctor.mjs to verify; doctor warns if the executor is missing.

Cheat Sheet & Best Practices

Frameworks: Polya: (1) Understand the problem (2) Devise a plan (3) Carry out the plan (4) Look back. IDEAL: Identify → Define context → Explore strategies → Act on best → Look back and learn.

Process: Be systematic: preview outcomes, sustain effort, examine from multiple angles, pace and self-monitor. Use diagrams, tables, patterns; break into subgoals; work backwards when useful.

Hacks: Number thoughts (thoughtNumber, totalThoughts); allow revision and branching. Start with an initial step count and refine. Use the MCP sequential-thinking tool for multi-step analysis and hypothesis verification when the full scope is unclear.

Certifications & Training

No cert. Problem-solving frameworks: Polya (Understand → Plan → Execute → Look back), IDEAL (Identify → Define → Explore → Act → Look back). Skill data: Systematic steps; preview outcomes; sustain effort; revise and branch; use diagrams/tables/patterns.

Hooks & Workflows

Suggested hooks: Optional: pre-plan or pre-complex-task hook to suggest sequential-thinking for multi-step tasks. Use with planner (primary) for planning and master-orchestrator when decomposing work.

Workflows: Use with planner (primary). Flow: complex task → load sequential-thinking → emit thoughts (MCP or executor) → revise → act. See core/router-decision for when router assigns planner.

Overview

This skill provides structured problem-solving through a flexible thinking process that can adapt and evolve. Each thought can build on, question, or revise previous insights.

Context Savings: ~97% reduction

• MCP Mode: ~15,000 tokens always loaded
• Skill Mode: ~500 tokens metadata + on-demand loading

When to Use

• Breaking down complex problems into steps
• Planning and design with room for revision
• Analysis that might need course correction
• Problems where the full scope isn't clear initially
• Multi-step solutions requiring maintained context
• Filtering irrelevant information
• Hypothesis generation and verification

Quick Reference

Use the MCP sequential thinking tool directly:

// Via MCP tool
mcp__sequential -
  thinking__sequentialthinking({
    thought: 'First, let me analyze the problem...',
    thoughtNumber: 1,
    totalThoughts: 5,
    nextThoughtNeeded: true,
  });

Tool: sequentialthinking

A detailed tool for dynamic and reflective problem-solving through thoughts.

Parameters

Key Features

• Adjust totalThoughts up or down as you progress
• Question or revise previous thoughts
• Add more thoughts even after reaching what seemed like the end
• Express uncertainty and explore alternatives
• Branch or backtrack (non-linear thinking)
• Generate and verify solution hypotheses

Process

1. Start with initial estimate of needed thoughts
2. Feel free to question/revise previous thoughts
3. Add more thoughts if needed, even at the "end"
4. Mark thoughts that revise or branch
5. Generate solution hypothesis when appropriate
6. Verify hypothesis based on Chain of Thought
7. Repeat until satisfied
8. Set nextThoughtNeeded: false only when truly done

Tool Execution

The sequential thinking tool is available via MCP. Use it directly in your responses:

// Execute a thinking step
mcp__sequential -
  thinking__sequentialthinking({
    thought: 'Your analysis here...',
    thoughtNumber: 1,
    totalThoughts: 5,
    nextThoughtNeeded: true,
  });

Configuration

MCP server configuration stored in config.json:

• Command: npx -y @modelcontextprotocol/server-sequential-thinking

Related

• Original MCP server: @modelcontextprotocol/server-sequential-thinking
• MCP Converter Skill: .claude/skills/mcp-converter/

Iron Laws

1. NEVER set nextThoughtNeeded: false before the solution hypothesis has been verified
2. ALWAYS adjust totalThoughts dynamically as problem complexity becomes clear
3. NEVER commit to the first approach without exploring at least one alternative via branching
4. ALWAYS mark corrections explicitly with isRevision: true and revisesThought: N
5. NEVER stop at the estimated thought count if the hypothesis remains unverified

Anti-Patterns

Memory Protocol (MANDATORY)

Before starting: Read .claude/context/memory/learnings.md

After completing:

• New pattern -> .claude/context/memory/learnings.md
• Issue found -> .claude/context/memory/issues.md
• Decision made -> .claude/context/memory/decisions.md

ASSUME INTERRUPTION: If it's not in memory, it didn't happen.