skvm-general

Drive the skvm CLI to profile models, AOT-compile skills, and run benchmarks.

2.5k|422|Updated Mar 5, 2026
One-click install
npx skills add https://github.com/openJiuwen-ai/jiuwenswarm --skill skvm-general
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: skvm-general
Source: https://github.com/openJiuwen-ai/jiuwenswarm/tree/main/jiuwenswarm/resources/agent/workspace/skills/skvm-general
Command: npx skills add https://github.com/openJiuwen-ai/jiuwenswarm --skill skvm-general

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It helps you efficiently use the SkVM toolchain to evaluate model capabilities, compile LLM skills ahead-of-time, and run or benchmark those skills without manually managing the workflow.

Core Features & Use Cases

  • Model profiling & caching: Profiles target models into a capability profile and reuses cached results to reduce repeated work.
  • AOT compilation & pipeline execution: Performs multi-pass AOT compilation to generate compiled skill variants suitable for a specific target model.
  • Run & benchmark workflows: Runs single ad-hoc tasks with a given skill and runs broader benchmarks to compare conditions such as original vs compiled variants, including jit proposal management.

Quick Start

Ask the AI: “Profile model <id>, then AOT-compile my skill at <path/to/SKILL.md>, and run one task from <path/to/task.json> using the compiled result.”

Frequently Asked Questions about skvm-general

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I AOT-compile an LLM skill for a specific target model?

To AOT-compile an LLM skill, you drive the skvm CLI to perform multi-pass compilation, generating skill variants tailored for a specific target model capability profile ahead of time.

What is model profiling and how does it help with skill execution?

Model profiling evaluates target models into a capability profile and caches results to reduce repeated work. This profile guides subsequent AOT compilation and skill execution workflows.

Do I need an OPENROUTER_API_KEY to run local proposal management commands?

No, you do not need an OPENROUTER_API_KEY for local proposal listing, showing, rejecting, or profile listing. The key is only required for LLM-calling commands like profile generation, aot-compile, pipeline, run, bench, and jit-optimize.

How can I benchmark original versus compiled skill variants across different models?

You benchmark original versus compiled skill variants by running broader benchmark workflows via the skvm CLI. This compares multiple execution conditions across one or more models to evaluate performance differences.

What prerequisites are needed to use the skvm toolchain?

To use the skvm toolchain, the skvm binary must be present on your system PATH. Additionally, an OPENROUTER_API_KEY environment variable is required for LLM-calling commands like profiling, compiling, and running tasks.

Can I run a single ad-hoc task using an already compiled skill variant?

Yes, you can run single ad-hoc tasks with a given skill using the compiled result. The skvm CLI drives this execution workflow, allowing you to validate a skill on one task without running a full benchmark.