Claude Skill
sentient-agi/EvoSkill
EvoSkill is an open-source Python framework that automatically discovers and synthesizes reusable agent skills from failed trajectories to boost coding agent performance.
Overview
Repository
Install this Skill
pip install -e .Registry
pip install -e .pip install harbor # Harbor (containerized benchmarks)pip install harbor # install the Harbor CLIpip install daytona
Summary
EvoSkill is an open-source framework that automatically discovers and synthesizes reusable agent skills from failed trajectories to improve coding agent performance.
EvoSkill — 一个开源框架，能够从失败的轨迹中自动发现并合成可复用的智能体技能，从而提升编码智能体的性能。
Key features
- Automatically discovers reusable skills from failed agent trajectories
- Synthesizes new skills to improve coding agent performance
- Open-source framework built in Python
- Designed for iterative agent improvement without manual intervention
Use cases
- Improving coding agent accuracy by learning from past failures
- Building a library of reusable skills for software development agents
- Automating agent skill discovery in continuous integration pipelines
- Enhancing autonomous code generation and debugging workflows
README excerpt
<div align="center"> <img src="./assets/evoskill_logo.png" alt="alt text" width="80%"/> <br> <h1>EvoSkill: Automated Skill Discovery for Coding Agents</h1> </div> <p align="center"> <a href="https://www.alphaxiv.org/abs/2603.02766"><img src="https://img.shields.io/badge/Paper-f73c6f?style=for-the-badge" alt="Paper"></a> <a href="https://www.sentient.xyz/blog/evoskill-automated-skill-induction-from-agent-failures"><img src="https://img.shields.io/badge/Blog-f73c6f?style=for-the-badge" alt="Blog"></a> <a href="https://sentient.xyz"><img src="https://img.shields.io/badge/Built%20by-Sentient%20Labs-f73c6f?style=for-the-badge" alt="Built by Sentient Labs"></a> <a href="https://x.com/SentientAGI"> <img src="https://img.shields.io/badge/-SentientAGI-grey?logo=x&style=for-the-badge"/> <a href="https://github.com/sentient-agi/EvoSkill/blob/main/LICENSE"><img src="https://img.shields.io/badge/License-Apache 2.0-007ec6?style=for-the-badge" alt="License: Apache 2.0"></a> </p> <b>Turn your general AI agents into state-of-the-art specialists with a benchmark and EvoSkill, a toolkit for automatically creating and improving AI skills, compatible with Claude Code, Codex CLI, OpenCode, OpenHands, Goose, Harbor, and more.</b> <b>EvoSkill</b> significantly extends the feedback-driven idea of <b>[GEPA](https://github.com/sentient-agi/gepa-plus)</b> from single-file optimization to complete agent evolution. Instead of only revising one prompt in place like GEPA, EvoSkill proposes multiple skill and prompt mutations jointly, evaluates new variants on held-out data, and has each iteration produce an entirely new agent program. <p align="center"> <img src="./assets/examples.png" alt="EvoSkill Architecture" style="width: 75%;"> </p> Install in seconds, then run `evo
Topics
No topics yet.