A sophisticated reasoning framework that applies the ReAct-Plan methodology to decompose complex tasks into structured, iterative analysis.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install deep-think
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install deep-think using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Deep Think (ReAct-Plan Framework)?
Deep Think is an advanced reasoning skill designed for AI agents to move beyond surface-level responses. It implements the ReAct-Plan framework, which interleaves chain-of-thought reasoning with explicit global planning, dynamic prediction, and continuous reflection. This approach helps the agent overcome myopic behaviors by maintaining a global view of the problem while executing local steps. By integrating this skill into Openclaw Skills, developers can ensure their agents handle complex, multi-layered queries with a level of systematic rigor that standard prompts cannot achieve.
Deep Think (ReAct-Plan Framework) Use Cases
- Systematic breakdown of multi-stage engineering or architectural problems.
- In-depth analysis of complex logic where requirements are not fully known at the start.
- Dynamic project planning where the agent must update its strategy based on intermediate findings.
- Strategic reasoning for mission-critical tasks requiring high-fidelity final synthesis.
How Deep Think (ReAct-Plan Framework) Works
- Activation: The skill triggers when it detects specific intent for deep analysis or systematic breakdown within Openclaw Skills.
- Initial Planning (P0): The agent establishes the core problem, success criteria, and a baseline execution plan (P0).
- Iterative Deep Dives: The agent performs 2 to 4 focused iterations. Each iteration focuses on a specific dimension, identifying new problems and updating the global plan.
- Dynamic Re-planning: During each dive, the agent reflects on new discoveries and only updates the parts of the plan that have changed.
- Final Synthesis: The process concludes with a comprehensive summary of core findings, the final refined plan, and an assessment of risks and success factors.
Deep Think (ReAct-Plan Framework) Setup
To utilize this skill, you must include the skill definition in your agent's configuration. It is triggered by phrases like "help me think deeply" or "provide a detailed breakdown."
# Add the skill to your local Openclaw Skills repository
openclaw install deep-think
Deep Think (ReAct-Plan Framework) Data Schema & Taxonomy
The skill organizes data using a structured Markdown taxonomy to ensure the reasoning path is transparent and easy to follow.
| Component | Description |
|---|---|
| P0 Plan | Includes Core Problem, Success Criteria, Knowns, and Unknowns. |
| Iterative Logic | Captures current step problem, concise analysis, and plan updates. |
| Synthesis | Final output containing 3-5 core insights and a risk mitigation strategy. |
| Plan Deltas | Optimized to show only changes in the plan rather than repeating the entire structure. |
Deep Think (ReAct-Plan Framework) Advanced Features
- Think Globally, Act Locally: Balances high-level strategy with granular task execution.
- Dynamic Prediction: Anticipates potential outcomes and obstacles before they occur during the reasoning phase.
- Belief State Reflection: Continuously assesses progress to update the agent's internal understanding of the problem.
- Conciseness Optimization: Specifically designed for Openclaw Skills to prevent context bloat by focusing only on critical findings and plan deltas.
SKILL.md
Loading
Related Openclaw Skills
Pywayne Data Structure
A specialized Python toolkit for managing hierarchical logic trees, union-find set operations, and seamless XML serialization.
Pywayne Crypto
An encryption and decryption toolkit for strings and bytes, featuring AES-128 support and Python code obfuscation.
Pywayne Cross Comm
A high-performance WebSocket service for real-time communication and file sharing between Python and other languages.
pywayne-bin-cmdlogger
A multi-threaded command-line utility that captures and logs all I/O streams in real-time while maintaining console interactivity.
pywayne-dsp
A specialized digital signal processing toolkit for Python designed to handle noise suppression, peak detection, and complex time-series analysis.
Pywayne Geometric Hull Calculator
A Python-based utility for calculating and visualizing geometric boundaries including convex hulls, concave hulls, and minimum bounding rectangles for 2D datasets.
Featured*
