Vibe Coding DiscoverSkillOpt
View on GitHubSkillOpt is a text-space optimizer that trains reusable natural-language skills for frozen LLM agents through trajectory-driven edits, validation-gated updates, and deployable best_skill.md artifacts.
SkillOpt optimizes natural-language skills for frozen LLM agents using rollout feedback, bounded edits, and held-out validation gates. It supports multiple model backends and coding-agent harnesses, and exports deployable best_skill.md artifacts.
Use Cases
Train reusable agent skills from scored task trajectoriesImprove coding-agent skills from past sessionsValidate skill edits against held-out tasksDeploy optimized skills as Markdown filesRun offline skill evolution for coding agents
Built With
- Language
- Python
- Frameworks
- OpenAI API · vLLM · Claude Agent SDK · Gradio
Tags
agent-skills · skill-optimization · self-evolving-agents · prompt-optimization · trajectory-learning · validation-gating · frozen-LLMs · offline-training