Microsoft open-sources SkillOpt, a framework for self-evolving agent skills
blaizedsouza · x · 2026-07-21
Microsoft open-sources SkillOpt, a self-evolving agent-skill framework
The post says Microsoft has open-sourced SkillOpt, a system for training agent skills the way you train a model, but without changing model weights.
What it claims:
- A base model performs the task.
- A separate optimizer evaluates the output and rewrites the skill instructions automatically.
- The system isolates successful paths, rejects edits that do not beat a benchmark, and transfers across different models.
- The attached README screenshot describes it as "Executive Strategy for Self-Evolving Agent Skills" and says it uses epochs, batch size, learning rates, validation gates, and bounded edit operations.
The pitch is that skill documents become the trainable state of a frozen agent, enabling reproducible improvement with zero inference-time model calls at deployment.
Related event: Microsoft Open-Sources SkillOpt for Self-Evolving Agent Skills(2 posts)→
More from coding & agent
- 'My intelligence stack': developer splits AI coding work across Opus 5.5, Fable and Astra with human review gates — sh_reya · 2026-10-27
- Baseten webinar: evaluate open models for coding agents by benchmarking on your own merged PRs — baseten · 2026-10-27
- Thorsten Ball on working with agents: run more evals, automate everything, use the best models — omarsar0 · 2026-10-27
- DeepMind paper: LLMs can't sample randomly on their own — pass random numbers from code instead — rohanpaul_ai · 2026-10-27
- BELL: Open-Source Stack Combines Browserbase, Eve, iMessage and Stripe Link for Personal Agents — jeff_weinstein · 2026-10-27
- Legora runs 2,000 parallel agent evals on Modal Sandboxes for legal AI — blelbach · 2026-10-27