Dynamically adjust AI reasoning depth based on task complexity to balance response speed and analytical depth.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install adaptive-reasoning
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 adaptive-reasoning 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 Adaptive Reasoning?
Adaptive Reasoning is a sophisticated pre-processing skill designed to optimize how AI agents handle diverse requests within the Openclaw Skills ecosystem. By evaluating incoming messages against specific complexity dimensions—such as multi-step logic, ambiguity, and architectural requirements—the skill determines whether to engage standard response modes or extended thinking capabilities. This ensures that high-stakes problems get the cognitive depth they require while routine tasks remain fast and cost-effective.
By implementing this skill, developers can provide a more intelligent user experience. Instead of a one-size-fits-all approach, the agent adopts a 'think-before-acting' mentality, surfacing visual indicators to signal the level of cognitive effort applied to a specific solution. This makes it a core component for anyone building high-performance agents using Openclaw Skills.
Adaptive Reasoning Use Cases
- Debugging complex distributed systems or identifying subtle race conditions.
- Designing multi-layered system architectures with conflicting constraints.
- Solving formal logic puzzles, mathematical proofs, or algorithmic challenges.
- Performing comprehensive security reviews on sensitive codebases.
- Reducing token consumption for routine lookups, greetings, and simple tasks.
How Adaptive Reasoning Works
- The skill intercepts every user message as a mental pre-processing step before generating a response.
- It calculates a complexity score from 0 to 10 based on weighted signals like multi-step logic, ambiguity, and novelty.
- Based on the calculated score, the skill selects a decision threshold: Fast, Standard, Reasoning, or Deep Thinking.
- If the threshold meets the criteria (score 6+), it activates extended reasoning using internal commands or status tools.
- The agent generates the response and appends a visual indicator (🧠 or 🧠🔥) if high-level reasoning was utilized.
- It automatically downgrades the reasoning level for subsequent simple messages to optimize performance and save tokens.
Adaptive Reasoning Setup
To integrate Adaptive Reasoning into your agent, add the logic to your skill configuration. Since this operates as a cognitive layer, it does not require external API keys. You can interact with it using these commands:
/reasoning on # Enable extended thinking for the current session
/reasoning off # Disable reasoning to prioritize speed and save tokens
/status # Check the current reasoning state of the agent
Adaptive Reasoning Data Schema & Taxonomy
The skill utilizes a weighted scoring matrix to evaluate every interaction. Data is organized based on the following taxonomy:
| Signal | Weight | Description |
|---|---|---|
| Multi-step logic | +3 | Planning, proofs, and debugging chains |
| Ambiguity | +2 | Nuanced questions and trade-offs |
| Code Architecture | +2 | System design and security reviews |
| High Stakes | +1 | Production changes or irreversible actions |
It manages internal state through a session_status object that tracks the reasoning boolean and the current complexity threshold (Fast <=2, Standard 3-5, Reasoning 6-7, Deep Thinking >=8).
Adaptive Reasoning Advanced Features
- Automatic token optimization via silent auto-downgrading when complexity drops.
- Visual feedback loops using emoji indicators to inform the user of processing depth.
- Manual overrides for power users to force specific reasoning states via CLI commands.
- Context-aware logic that respects explicit user requests for speed or brevity.
- Seamless integration as a pre-processing middleware for any Openclaw Skills workflow.
SKILL.md
Loading
Related Openclaw Skills
OpenCode Remote Control
A sophisticated remote management skill for OpenCode servers that enables multi-session orchestration and automated progress monitoring via HTTP API.
LiveAvatar
LiveAvatar provides a real-time video and voice interface for your agents, enabling face-to-face AI communication.
Kenya Tax Rates Skill
A specialized utility for calculating accurate Kenya payroll deductions, including PAYE, SHIF, and Housing Levy using 2024/2025 rates.
Agent Money Tracker
A comprehensive TypeScript library for AI agents to manage finances, budgets, and savings through natural language parsing and structured data.
Pattern Analyst
An AI agent skill designed to observe interactions, identify behavioral patterns, and surface insights that align daily actions with long-term goals.
Scamper
A structured ideation skill that applies the SCAMPER framework to expand concepts, adapt frameworks, and unblock creative workflows.
Featured*
