Stats
Actions
Tags
From autoimprove
Main entry point for the autonomous improvement loop. Use when the harness calls `Skill(autoimprove)`, when the user runs `/autoimprove`, or when the user asks to start the full research → experiment → judge → converge flow. This is an alias for the `run` skill. It exists so callers can invoke the top-level `autoimprove` skill name directly without failing with "Unknown skill: autoimprove".
How this skill is triggered — by the user, by Claude, or both
Slash command
/autoimprove:autoimprove [--experiments N] [--theme THEME] [--resume] [--phase propose][--experiments N] [--theme THEME] [--resume] [--phase propose]This skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Treat this invocation as equivalent to the `run` skill.
Treat this invocation as equivalent to the run skill.
Before doing any work, read skills/run/SKILL.md, then follow its instructions exactly while preserving any user-supplied arguments from this autoimprove invocation.
Key requirements:
skills/run/SKILL.md.run.run skill after loading it.skills/run/SKILL.md. If you find yourself writing steps here, move them to run instead.--experiments 5 --theme lint_warnings, forward both flags verbatim when executing the run flow. Do not interpret or discard arguments at this layer.run skill somehow delegates back to autoimprove, stop and report a routing loop.autoimprove. If you are already inside the run flow, stay there — do not re-enter via this alias.report, status, or rollback skills respectively.npx claudepluginhub tokyo-megacorp/autoimproveCreates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.