AgentSkillsCN

ultrathink

针对多步骤问题解决的顺序思维CLI工具。适用于将复杂问题拆解为若干步骤、规划实施方案、调试多层问题，或在需要清晰的逐步推理、自信追踪与假设管理时的场景。

SKILL.md
--- frontmatter
name: ultrathink
description: Sequential thinking CLI for multi-step problem solving. Use when breaking down complex problems into steps, planning implementations, debugging multi-layered issues, or when explicit step-by-step reasoning with confidence tracking and assumption management is beneficial.

UltraThink - Sequential Thinking CLI

A tool for dynamic and reflective problem-solving through structured, sequential thoughts.

When to Use

Use this skill for:

  • •Breaking down complex problems (>3 reasoning steps)
  • •Planning and design with room for revision
  • •Architecture decisions with multiple trade-offs
  • •Algorithm design and optimization
  • •Debugging multi-layered issues
  • •Analysis that might need course correction

Do NOT use for:

  • •Simple one-step answers
  • •Straightforward code edits
  • •Basic file operations
  • •Simple factual questions

Instructions

  1. •Start a session with initial thought and estimated total
  2. •Continue session using session_id from response
  3. •Use --confidence to track uncertainty (0.0-1.0)
  4. •Revise thoughts with --is-revision --revises <num>
  5. •Branch to explore alternatives with --branch-from <num> --branch-id <name>
  6. •Track assumptions with --assumptions, --depends-on, --invalidates
  7. •Monitor risky_assumptions in response
  8. •Continue until next_thought_needed is false

Quick Start

bash
# Start new session
uv run ${CLAUDE_PLUGIN_ROOT}/skills/ultrathink/scripts/ultrathink.py -t "Analyzing the problem..." -n 5

# Continue session
uv run ${CLAUDE_PLUGIN_ROOT}/skills/ultrathink/scripts/ultrathink.py -t "Next step..." -n 5 -s <session-id>

# With confidence
uv run ${CLAUDE_PLUGIN_ROOT}/skills/ultrathink/scripts/ultrathink.py -t "I believe X is correct" -n 3 -c 0.8

# With assumption
uv run ${CLAUDE_PLUGIN_ROOT}/skills/ultrathink/scripts/ultrathink.py -t "Assuming Redis available" -n 3 \
  --assumptions '[{"id":"A1","text":"Redis is available"}]'

Example Session

code
Thought 1 (confidence: 0.6): "Analyzing access patterns for caching strategy..."
Thought 2 (confidence: 0.7): "Two approaches: LRU or LFU..."
Thought 3 (revision of 2): "Should also consider TTL-based expiration..."
Thought 4 (branch from 2): "Exploring hybrid LRU+TTL approach..."
Thought 5 (confidence: 0.95): "Hybrid approach recommended."

Reference

For complete documentation including all CLI options, response schema, and advanced examples, see references/REFERENCE.md.