sentient-agi/EvoSkill — GitHub trending stats & insights

Featured

sentient-agi/EvoSkill

#AI agent #AI coding assistant #AI skills

EvoSkill — An open-source framework that automatically discovers and synthesizes reusable agent skills from failed trajectories to improve coding agent performance.

Visit GitHub

Python

821

10 contributors

Apache License 2.0

Social mentions

Recent discussions about this repository across the web

EvoSkill v1.2.0 just got a major upgrade with Harbor integration. Harbor is a framework built for evaluating AI agents on containerized benchmark tasks. With this integration, EvoSkill can now train…

@Krypto_Kratos · x.com

Before: → evolution loops on static questions Now: → SWE-bench Verified → Terminal-Bench 2.0 → Aider Polyglot → 185+ more benchmark environments Static evals → real-world coding verification.

@smoothasfkk · x.com

This is no longer just “run benchmark → get score.” It’s starting to look like continuous large-scale agent training infrastructure for real coding systems What stands out to me most is the scale…

@Flintex6 · x.com

What makes EvoSkill interesting isn’t just the benchmark gains. It’s the direction. The broader Sentient research path is pointing toward systems that can improve: • prompts • skills • memory •…

@connectwithveee · x.com

Social mentions

Repository activities