GitHub Repository

Code execution service for AI agent skills. Runs in isolated subprocesses with skill caching.

6 starsGo

Skills RCE – Code execution server for AI agent skills

Name: Skills RCE – Code execution server for AI agent skills
Availability: InStock
Author: muxidev

by muxidev·Mar 9, 2026·1 point·0 comments

Visit Project View on HN

AI Analysis

●●SolidShip It

Single Go binary with OpenAPI spec beats framework-locked orchestration platforms.

Strengths

•Skill directory caching means upload once, execute many times efficiently
•Docker and Singularity images bundle Python, Node, Bun, and Go runtimes
•Bearer token auth and configurable timeouts for production deployment

Weaknesses

•Code execution sandboxes already solved by E2B, Modal, and Replicate
•No built-in rate limiting or resource quotas beyond timeout configuration

Post Description

Hey HN

We've built this as part of MUXI, an open-source AI agent infrastructure platform.

The problem: agents need to execute code, not just generate it. Most solutions either give full shell access (terrifying) or sandbox so heavily that nothing useful runs.

skills-rce sits in the middle: a single binary with a standardized OpenAPI-spec'd REST API. You point it at a skill, it runs the code in an isolated subprocess with skill directory caching, and returns the result. Language-agnostic, single endpoint.

It ships as a single binary, but we recommend running it via Docker so execution never touches your host machine. The image is ready to go.

It's intentionally simple. No orchestration opinions, no framework lock-in. Just a code execution service your agents can call.

Apache 2.0. Feedback welcome!