Skip to content
Get packs

SkillOpt-Sleep Self-Improvement

Lets Codex review past sessions overnight and propose validated edits to its skills and memory

Source Download skill
Repository
microsoft/SkillOpt
Stars
18,813
Licence
MIT
Files
1 file

What it does

This skill teaches a Codex coding agent to run a background 'sleep' cycle. It reads your past local Codex sessions, finds recurring tasks, re-runs them through a chosen backend, and proposes bounded edits to a configured skill and the project's CLAUDE.md. Changes are staged for review, and with the default gate only edits that improve a held-out validation score are kept. Live files change only when you explicitly adopt them or request auto-adopt.

Great for

  • Developers who use Codex heavily and want its project instructions and skills refined from real usage
  • Teams experimenting with scheduled, offline self-optimization of agent skills
  • People comfortable running shell scripts and reviewing staged proposals before adoption

Skip it if

  • Users who want a quick one-off answer, since the cycle can take time and consume model budget when using the codex, claude, or azure_openai backends (the mock backend spends no API budget)
  • Anyone unwilling to review staged changes, because auto-adopt can overwrite live skill and CLAUDE.md files (backups are kept for existing targets)
  • Sensitive projects where past session transcripts should not be read or sent to model backends, since it reads local transcripts and may send them to external services

Try asking

  • “Run a dry-run of SkillOpt-Sleep against my recent Codex sessions using the mock backend before anything else”
  • “Harvest my past Codex sessions and show me the recurring tasks it finds for my example skill”
  • “Schedule a nightly Codex sleep cycle at 3:17 for this project and stage proposals for review”
  • “Show the status of my last SkillOpt-Sleep run and adopt only the proposals for the skill I selected”
Step-by-step guide: how to use SkillOpt-Sleep Self-ImprovementInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 1 original files, the MIT licence and an attribution note.

    skillopt-sleep.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip skillopt-sleep.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip skillopt-sleep.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Always start with a dry-run using --backend mock and check status before a real run with --backend codex.
  • Set an explicit --target-skill-path to a .agents skill so the result is visible to Codex, and select skills with --skill NAME rather than adopting everything.
  • Review staged proposals under <project>/.skillopt-sleep/staging before adopting, and treat accepted held-out gains as run-specific rather than guaranteed.

What's inside

  • SKILL.md

Redistributed unmodified from microsoft/SkillOpt under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.