Skillopt Sleep

This gives your local Claude agent a nightly self-improvement cycle. It reads your past sessions, extracts recurring tasks, replays them offline against a held-out validation set, and only keeps memory or skill edits that measurably improve performance. The whole thing stages proposals first so nothing changes until you explicitly adopt them. Default mode is mock replay with zero API spend, which is enough to see the mechanics work. Switch to a real backend and it burns your API budget to genuinely get better at your specific workflows. If you've ever wanted proof that distilling experience into durable competence actually works, run the bundled experiment script, it shows the held-out score rise and the gate block bad edits.

skills.sh installs: 241
GitHub stars: 19K
last updated: Jun 10, 2026

View on GitHub

Install

$ npx -y skills add microsoft/skillopt --skill skillopt-sleep --agent claude-code

Installs into .claude/skills of the current project.

Files

SkillOpt-Sleep: usage-driven self-evolution for a local Claude agent

SkillOpt-Sleep gives the user's agent a sleep cycle. On demand or on a nightly schedule, it reviews real past Claude Code sessions, re-runs recurring tasks through the selected backend, and consolidates what it learns into memory (CLAUDE.md) and skills (SKILL.md). With the default validation gate enabled, it keeps only changes that improve a held-out score. Live files change only through explicit adoption or a user-requested --auto-adopt. It aims to improve this user's recurring work, while making each accepted proposal measurable on the run's held-out tasks, with no model-weight training. It is the deployment-time analogue of training: short-term experience → long-term competence.

SKILL.md on GitHub