A 4-step structured thinking framework that enforces systematic reasoning to significantly improve the output quality of any LLM.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install llmbooster
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 llmbooster 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 LLMBooster?
LLMBooster is a specialized thinking framework designed to solve the common pitfalls of low-end LLMs, such as jumping to conclusions or overlooking critical details. Unlike automated scripts that rely on external API calls, this Openclaw Skills contribution enables the LLM to follow a self-correcting cognitive process. By breaking down tasks into discrete stages, it ensures the model doesn't just generate text, but actively reasons through the problem.
At its core, LLMBooster forces a transition from raw generation to structured engineering. It leverages a Plan-Draft-Critique-Refine pipeline, making it an essential addition to Openclaw Skills for developers who require high-fidelity technical documentation, complex code reviews, or in-depth analytical reports from their AI agents.
LLMBooster Use Cases
- Improving output quality for low-end or smaller LLM models that tend to produce shallow responses.
- Generating high-stakes technical documentation and complex code reviews.
- Conducting in-depth competitive analysis and multi-step decision support.
- Ensuring comprehensive coverage in lengthy or multifaceted research tasks.
- Refining rough drafts into polished, professional-grade content through automated self-critique.
How LLMBooster Works
- The skill is triggered by specific keywords or the /booster command within the Openclaw Skills environment.
- The LLM initializes the framework by reading
plan.mdto create a structured outline of the task. - It proceeds to the Draft phase using
draft.mdto generate a full, detailed initial response. - The Self-Critique phase is activated via
self_critique.md, where the LLM identifies weaknesses, missing details, or logical errors in its own draft. - Finally, the LLM applies all identified improvements during the Refine phase to deliver a polished final output.
LLMBooster Setup
To integrate LLMBooster into your local environment, ensure you have the Openclaw Skills directory structure ready. Installation involves placing the skill files into your workspace and utilizing the CLI for state management.
# Navigate to your Openclaw Skills workspace
cd ~/.openclaw/workspace/skills/llmbooster
# Enable the booster framework
/booster enable
# Verify status and configuration
/booster status
LLMBooster Data Schema & Taxonomy
LLMBooster manages its lifecycle through a structured metadata and prompt-based taxonomy. This ensures that every Openclaw Skills interaction is logged and optimized.
| Component | File / Path | Description |
|---|---|---|
| Core Logic | booster.py |
Handles the pipeline execution and helpers. |
| State Management | state_manager.py |
Tracks usage statistics and current depth settings. |
| Prompt Templates | prompts/*.md |
Contains the logic for Plan, Draft, Critique, and Refine steps. |
| Configuration | config.schema.json |
Defines the parameters for thinking depth and visual feedback. |
LLMBooster Advanced Features
- Dynamic Thinking Depth: Adjust the pipeline from 1 (Plan only) to 4 (Full pipeline) based on the required speed vs. quality tradeoff.
- Visual Progress Streaming: Real-time feedback in the CLI showing the status of each reasoning step (Plan, Draft, Critique, Refine).
- Command Line Integration: Full control over the framework using
/boostercommands for stats, depth adjustment, and toggling. - No External Dependencies: Operates entirely within the LLM's reasoning context without needing additional API endpoints or complex backend infrastructure.
- Usage Analytics: Track tasks processed and last-used timestamps via the stats command.
SKILL.md
Loading
Related Openclaw Skills
Sports Arbitrage Scanner
An automated scanner that identifies sports betting arbitrage opportunities to lock in guaranteed profits regardless of match outcomes.
Scribe Docs
An autonomous writing and copywriting engine designed to eliminate the blank page and deliver deployable, high-conversion drafts instantly.
AgentMM
AgentMM provides AI agents with persistent, structured memory storage and high-fidelity logging capabilities to improve long-term recall and debugging.
AI Video Generation Skill
An end-to-end pipeline to transform text prompts into full-length videos including image generation, synthesis, voice-overs, and automated editing.
Quentn Integration
A powerful integration for Quentn that enables AI agents to manage marketing automation, CRM data, and customer workflows seamlessly.
Multicall
A powerful CLI utility for analyzing, tracking, and auditing blockchain multicall operations and protocol security.
Featured*
