An automated complexity-detection tool that switches between Claude Haiku and Sonnet to balance speed and reasoning power.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install model-switcher
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 model-switcher 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 Model Switcher?
Model Switcher is an essential utility for developers using Openclaw Skills to optimize their AI agent workflows. It functions as a dynamic router that evaluates the complexity of a user's prompt—detecting needs for refactoring, architectural design, or deep analysis—and automatically shifts the session to the most appropriate model. This ensures that powerful resources like Claude Sonnet are utilized only when necessary, while faster models like Haiku handle routine queries, resulting in a more responsive and cost-effective development experience.
Model Switcher Use Cases
- Scaling model power during deep architectural analysis or system design tasks.
- Reducing latency by using lightweight models for simple code queries or documentation lookups.
- Optimizing token usage and cost by dynamically matching model intelligence to task difficulty.
- Transitioning between models automatically during complex debugging sessions without manual configuration.
How Model Switcher Works
- The agent intercepts the user message and scans for specific intent or complexity keywords.
- The current session state is queried using the session status tool to identify the active model.
- If a high-complexity task is detected while on a fast model, the agent triggers a switch to the powerful model.
- The agent provides a brief notification to the user about the model change to ensure transparency.
- Once the complex operation is complete, the agent can optionally switch back to the default fast model for subsequent interactions.
Model Switcher Setup
To integrate this capability into your Openclaw Skills setup, ensure your session status tool is configured to recognize your specific model endpoints. Use the following commands within your agent's execution logic:
# Trigger a switch to the high-performance sonnet model
session_status({ model: "kiro-cli" })
# Revert to the high-speed haiku model
session_status({ model: "haiku" })
Model Switcher Data Schema & Taxonomy
The skill manages model state transitions through the following data structures:
| Parameter | Type | Description |
|---|---|---|
| model | string | The identifier for the target model (e.g., haiku, kiro-cli, or default). |
| session_status | tool | The primary interface for reading and writing model configuration state. |
| trigger_list | array | A collection of keywords that signify a need for high-reasoning capabilities. |
Model Switcher Advanced Features
- Context-aware switching that maintains high-reasoning models throughout a continuous complex discussion.
- Support for multi-lingual keyword detection, optimized for both English and Chinese technical terminology.
- Batch processing logic to group related complex tasks together on a powerful model before switching back.
- Explicit manual override support allowing users to mention model names to trigger immediate switches.
SKILL.md
Loading
Related Openclaw Skills
WeChat Article Reader
A production-grade tool for crawling and parsing WeChat Official Account articles with high precision and automated content cleaning.
Reddit Quote Topaz Carousel Creator
Automates the creation of high-fidelity Instagram carousels by pairing curated Reddit travel quotes with Topaz Labs AI-upscaled imagery.
Reddit Quote Carousel
Transform Reddit-sourced travel recommendations into professional Instagram carousel posts automatically.
Lead Scorer
An automated lead qualification engine that scores domains based on technical signals, content velocity, and business legitimacy.
Memic Context Engineering SDK
A managed context engineering platform that provides RAG, Text2SQL, and hybrid search to ground AI agents in real-world data with high efficiency.
Usage Tracker Skill
A robust utility for logging AI model consumption and generating detailed cost reports across multiple API providers.
Featured*
