adaptive-reasoning

What This Skill Does

The adaptive-reasoning skill acts as the intellectual engine for your OpenClaw agent, providing a dynamic framework for assessing task complexity before generating a response. Rather than treating every query with the same level of depth, this skill implements a sophisticated scoring system that evaluates incoming requests based on multi-step logic, ambiguity, technical architecture, mathematical difficulty, and overall impact. By assigning weights to these dimensions, the skill decides whether the agent should operate in 'Fast Mode' for simple inquiries or trigger 'Extended Thinking' (Reasoning Mode) for complex problems that require deep analysis or structural planning.

Installation

To integrate this intelligence layer into your OpenClaw environment, execute the following command in your terminal or command-line interface:

clawhub install openclaw/skills/skills/enzoricciulli/adaptive-reasoning

Once installed, the skill runs automatically as a pre-processing step for every user message. You can manage the state of the reasoning engine manually using the /reasoning on and /reasoning off commands or check the state with /status.

Use Cases

Adaptive-reasoning is best suited for environments where the agent faces a high variance in query types. It excels in:

• Software Development: Distinguishing between a quick syntax check and a high-level system architecture design review.
• Complex Problem Solving: Providing the necessary 'brainpower' for debugging distributed systems or creating detailed logic proofs.
• Strategic Planning: Determining when to take the time to map out project steps versus answering simple status updates.
• Cost Efficiency: Ensuring that token consumption is optimized by disabling extended thinking for routine or repetitive tasks that do not require deep analysis.

Example Prompts

1. "How do I write a basic loop in Python?" (The skill will identify this as a routine task and respond immediately without overhead.)
2. "Compare the pros and cons of using a microservices architecture versus a monolith for a high-traffic e-commerce platform." (The skill will trigger reasoning mode to perform a thorough, multi-factor analysis.)
3. "Analyze the following error logs and propose a multi-step debugging plan to resolve the race condition." (The skill will activate deep thinking mode, signified by the 🧠🔥 indicator.)

Tips & Limitations

• Manual Override: Always use /reasoning off if you are involved in a time-sensitive conversation where speed is the absolute priority over nuance.
• Context Awareness: The skill is designed to respect explicit user instructions. If you include keywords like 'quick', 'tldr', or 'just give me the answer', the adaptive engine will bypass the extended reasoning logic regardless of the inherent complexity of the task.
• Token Management: While deep thinking leads to better accuracy on complex topics, it does increase the token count; use it wisely in long-running sessions.