A specialized CLI for profiling model capabilities and AOT-compiling agent skills for optimized cross-model execution.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install skvm-general
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 skvm-general 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 SkVM General Usage?
SkVM General is the primary command-line interface for the Skill Virtual Machine, a system designed to AOT-compile and execute LLM agent skills across heterogeneous model architectures. It enables developers to bridge the gap between varying model capabilities by transforming standard skill definitions into optimized versions that are specifically tailored to a target model's profile. This tool is essential for anyone looking to deploy Openclaw Skills that remain robust and efficient regardless of whether the underlying LLM is a frontier model or a smaller, specialized engine.
SkVM General Usage Use Cases
- Profiling target models to map 26 primitive capabilities into a Target Capability Profile (TCP).
- Performing Ahead-of-Time (AOT) compilation of agent skills to handle capability substitution and compensation.
- Running ad-hoc debugging tasks to validate skill logic and adapter compatibility before deployment.
- Benchmarking Openclaw Skills against baselines to measure the impact of JIT and AOT optimizations.
- Reviewing and accepting optimization proposals to iteratively improve the performance of agentic workflows.
How SkVM General Usage Works
- The system performs a prerequisite check to ensure the skvm binary is present and an API key is available for LLM-calling commands.
- A Target Capability Profile is generated for the specific LLM, recording its technical strengths and weaknesses across various operational primitives.
- The AOT compiler runs up to three passes to rewrite the skill, generate environment-binding scripts, and extract workflow parallelism.
- Developers run the skill against specific tasks or full benchmarks to verify performance gains in different environments.
- Optimization artifacts are stored as proposals, which can be inspected and then explicitly accepted to deploy the improved Openclaw Skills.
SkVM General Usage Setup
To install the skvm CLI, run the following command in your terminal:
curl -fsSL https://skillvm.ai/install.sh | sh
Ensure your API key is configured for LLM-based operations such as profiling or compiling:
export OPENROUTER_API_KEY="your_key_here"
SkVM General Usage Data Schema & Taxonomy
The skill manages data through a structured local filesystem hierarchy to ensure reproducibility and caching:
| Data Type | Default Path | Description |
|---|---|---|
| Profiles | .skvm/profiles/ |
Cached Target Capability Profiles (TCP) for LLMs |
| Proposals | .skvm/proposals/ |
Stored optimization rounds and metadata for AOT and JIT results |
| Benchmarks | .skvm/log/bench/ |
Detailed logs and session data from performance evaluations |
| Cache | ~/.skvm |
Global runtime cache for system artifacts |
SkVM General Usage Advanced Features
- Three-pass AOT compilation involving gap analysis, dependency manifest generation, and DAG parallelism extraction.
- Concurrent execution support for profiling and benchmarking across multiple models simultaneously.
- Detached optimization mode for long-running JIT processes with status tracking and remote cancellation.
- Flexible adapter support for various frameworks including openclaw, opencode, and hermes.
- Post-run asynchronous judging to minimize latency during large-scale benchmarking sessions.
SKILL.md
Loading
Related Openclaw Skills
Video Create AI
An AI-driven video production skill that transforms images, clips, and text prompts into professional rendered videos using remote GPU nodes.
In Video
A cloud-powered AI video editor that turns natural language instructions into fully rendered 1080p video clips without manual sliders.
Editor Modi: AI-Powered Video Editing & Export
Editor Modi is an AI-driven video editing skill that automates cutting, transitions, and professional exports without requiring local software or GPU power.
In Video Free AI Video Editor
A professional-grade AI video editing skill that transforms raw footage into polished, watermark-free videos via simple text commands.
Generator In Video
A cloud-powered AI video editor that transforms raw footage into professional highlight reels using simple text descriptions.
Music AI
An automated AI video production tool that transforms audio tracks into beat-synced music videos using cloud rendering.
Featured*
