# microsoft/SkillOpt

> A text-space optimizer that trains reusable natural-language skills for frozen LLM agents via trajectory-driven edits, validation-gated updates, and a deployable best_skill.md artifact. Model weights stay untouched.

- Magnitude: 3.6 out of 10 — Breakout
- Stars: 17,743 total · +49 stars today, ≈ 95 by evening
- Star trust: star growth looks organic
- Category: AI agents · Language: Python · License: MIT · Created: 2026-05-08 · Last push: 2026-09-05
- GitHub: https://github.com/microsoft/SkillOpt · Homepage: https://aka.ms/skillopt · Page: https://gitnova.dev/en/r/microsoft/SkillOpt

## Useful for

- Train an agent skill on your own dataset and get best_skill.md without fine-tuning the model
- Run nightly self-evolution via skillopt-sleep for Claude Code or Codex
- Evaluate transfer of a trained skill across different models and harnesses

## Why it’s here

- 49 stars so far today, about 95 expected by the end of the day. The usual pace is 40 per day, so that's 2.4× as much.
- Over the last two days the pace is 2.6× that of the previous week and a half.
- The spike has held for 3 days in a row — not a one-off blip.
- GitHub Trending Python today: #6, +111 stars.
- About 13 forks a day — people are taking the code.

## Star trust

Star growth looks organic.

Star-trust labels are heuristics based on the repository’s behavior, not a check of every stargazer.

## Numbers

- Forks: 1,658
- Issues and pull requests: 292
- Watchers: 69
- Average over the last week: 69 per day
- Usual pace: 40 per day
- Stars in the last hour (measured): 6
- Latest release: v0.2.0 (2026-07-02)

## Stars per day, last 30 days (oldest → newest, today is partial)

2026-08-30 … 2026-09-28: 38, 50, 47, 41, 26, 22, 14, 26, 37, 28, 28, 37, 57, 51, 70, 59, 51, 36, 37, 43, 32, 28, 33, 51, 40, 54, 39, 91, 112, 49

## Spotted in now

- GitHub Trending Python today: #6, +111 stars

## Similar by description

1. **flybook-git/flm** — 0.0 · Quiet · Science & research · Python · +0 stars today, ≈ 0 by evening
   A research project that trains a small adapter on top of a frozen LFM2.5-1.2B LLM; the adapter reads the state of the full MaleCNS v1.0 fly connectome (166,700 nodes, 25.5M edges) and adjusts next-token logits. Only the adapter is…
   Full card: https://gitnova.dev/en/r/flybook-git/flm.md
2. **microsoft/agent-lightning** — 1.9 · Steady · AI agents · Python · +2 stars today, ≈ 10 by evening
   A lightweight (~3,500 LOC) reinforcement-learning framework for training AI agents with their real harnesses: agents run unchanged while a proxy gateway captures training data. Includes Kubernetes support and coding-agent training examples.
   Full card: https://gitnova.dev/en/r/microsoft/agent-lightning.md
3. **Human-Agent-Society/reef** — 4.6 · Peaking · AI agents · Python · +196 stars today, ≈ 389 by evening
   Continual learning infrastructure for self-improving agents: connects agent inference, feedback collection, learning, and versioned delivery of updates. Lets you train model weights or improve an agent's harness (prompts, rules, skills).
   Full card: https://gitnova.dev/en/r/Human-Agent-Society/reef.md

---

Magnitude (0–10) measures how fast and how unusually interest in a repository is growing right now. It is not a quality score.
Days are UTC. “So far today” is a fact; “expected by the end of the day” is a forecast.
Summaries and use cases are written by an LLM (DeepSeek V4.1 Flash) from the README and may be inaccurate: verify specific claims (benchmarks, speed, hardware) in the repository itself.
Data as of 2026-09-28 13:02 UTC, updated every 30 minutes.