JimmyMa99

JimmyMa99 / TreeSkill

Public

Skill optimization framework for LLMs — evolve system prompts via textual gradient descent with beam search, human-in-the-loop annotation, and DPO preference data export.

10
1
100% credibility
Found Mar 23, 2026 at 10 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

TreeSkill is a framework for evolving AI agent prompts using user feedback and textual gradient descent without model training.

How It Works

1
📥 Discover TreeSkill

You find this helpful tool that makes AI assistants smarter without any complicated training.

2
🔧 Get it ready

Download and launch it easily so your personal AI helper is up and running in moments.

3
🔗 Connect your AI

Link it to a smart AI service like you would connect a new app, keeping everything private and secure.

4
🧠 Create or pick an assistant

Choose a ready-made helper or make your own for writing, classifying papers, or any task you need.

5
💬 Chat and give feedback

Talk to your assistant, and when it's not quite right, simply say what you'd like better to guide it.

6
✨ Watch it improve

Tell it to get better, and it learns from your notes to become more helpful and accurate right away.

🎉 Your smart assistant evolves

Over time, your AI gets personalized and sharper, handling tasks just the way you want without extra work.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 10 to 10 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is TreeSkill?

TreeSkill is a Python framework that evolves LLM system prompts into optimized agent skills using textual gradient descent and beam search, all without model training or GPUs—just API calls. You interact via CLI with human-in-the-loop annotation commands like /bad, /rewrite, or /optimize, building hierarchical skill trees that split, prune, and graft for tasks like paper classification or writing assistance. It supports OpenAI, Anthropic, and compatible APIs, plus tools via Python scripts, HTTP endpoints, or MCP servers, and exports DPO preference data.

Why is it gaining traction?

It stands out by automating all skill optimization from feedback traces, with beam search over gradient templates for stable prompt evolution—far beyond manual tweaking in github skill anthropic or github skill claude setups. Resumeable sessions and skill tree optimization handle complex behaviors like skill mix optimization or skill set optimization reinforcing language model behavior via transferable skills. Developers dig the train-free path to github skill tree agents, complete with annotation workflows and DPO export.

Who should use this?

AI agent builders crafting domain-specific prompts, like optimization skill for github skill copilot extensions or pathfinder heal skill optimization in RPG tools. Teams doing human-in-the-loop annotation on datasets for paper classification or writing skills. Prompt engineers exploring skill tree optimization arc raiders-style hierarchies without fine-tuning overhead.

Verdict

Promising for github skill tree experiments at 10 stars and 1.0% credibility—strong demos and docs outweigh light tests and maturity. Prototype it for your next LLM agent if feedback-driven optimization skill appeals.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.