A comprehensive reasoning framework that injects deep thinking, systematic verification, and transparent confidence scoring into AI agents.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install thinking
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 thinking 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 Thinking Protocol Skill?
The Thinking Protocol is a sophisticated skill designed to transform how agents process information within Openclaw Skills. Inspired by the Thinking-Claude project, it forces the AI to move beyond surface-level responses by implementing an organic inner monologue. This ensures that the agent builds understanding progressively, acknowledges its own reasoning flaws, and corrects mistakes in real-time before providing a final answer.
Version 2.1 introduces a rigorous Confidence Score System and Uncertainty Declaration. By quantifying reliability through specific factors like evidence quality and reasoning clarity, this skill provides a layer of transparency and trust that is essential for complex technical or analytical tasks. It effectively turns a standard AI into a self-correcting, highly analytical collaborator.
Thinking Protocol Skill Use Cases
- Executing complex technical tasks where performance, security, and scalability are critical.
- Performing high-risk operations such as financial analysis or system configuration changes that require a verification protocol.
- Conducting deep research and analysis where distinguishing between verified facts and assumptions is vital.
- Creating high-quality technical content that demands a consistent tone and logical flow.
How Thinking Protocol Skill Works
- Initial Engagement: The agent rephrases the query to ensure comprehension and maps known vs. unknown elements.
- Problem Analysis: The core request is broken down into implicit and explicit requirements and constraints.
- Hypothesis Generation: Multiple interpretations and solutions are considered simultaneously to avoid premature commitment.
- Natural Discovery: The agent uses a stream-of-consciousness thinking block to identify patterns and question initial assumptions.
- Systematic Verification: A secondary check is triggered for high-risk recommendations or complex reasoning chains to ensure logical consistency.
- Confidence Assessment: The final response is appended with a detailed confidence score and uncertainty tags to provide full transparency to the user.
Thinking Protocol Skill Setup
To install the Thinking Protocol, copy the skill directory into your local workspace for Openclaw Skills:
cp -r skills/thinking ~/.openclaw/workspace/skills/
Thinking Protocol Skill Data Schema & Taxonomy
The skill utilizes a structured reasoning and reporting format to maintain transparency throughout the agent's lifecycle:
| Component | Format | Description |
|---|---|---|
| Inner Monologue | thinking code block |
A natural language stream-of-consciousness process. |
| Uncertainty Tags | {uncertain}, {assumption}, etc. |
Inline markers for specific types of non-verified info. |
| Confidence Assessment | Markdown Table | A 0-100% score based on evidence, clarity, and expertise. |
| Verification Triggers | Logic Hooks | Conditions like "Conclusion Change" that force a re-evaluation. |
Thinking Protocol Skill Advanced Features
- Automatic Verification Triggers: Initiates an internal audit if high-risk recommendations or contradictions are detected.
- Domain-Specific Profiles: Tailors thinking logic for technical problems, content creation, or research-heavy tasks.
- Confidence Factor Scoring: Breaks down reliability into Evidence Quality, Reasoning Clarity, Domain Expertise, and Information Completeness.
- Auto-Correction Mechanism: Explicitly flags errors in previous reasoning and shows the corrected path to the user.
SKILL.md
Loading
Related Openclaw Skills
CFP Financial Planner
A professional-grade financial planning assistant that applies the six pillars of the CFP methodology to your personal wealth management.
Prompt Creator HE
A specialized AI assistant for creating professional, cinematic, and culturally authentic image and video prompts in both Hebrew and English.
core-executing-plans
A specialized execution framework for AI agents to process batch implementation plans with built-in human verification checkpoints.
Pharmacist
A professional pharmacy consultant skill for medication information, drug safety, and OTC treatment advice.
RAGFlow Dataset Ingest
A comprehensive automation suite for managing RAGFlow datasets, document parsing, and high-precision information retrieval.
X Thread Reader
A utility to convert X/Twitter threads into consolidated text or PDF documents for easier reading and archiving.
Featured*
