Run a configurable loop that incrementally improves a target (agent, skill, doc, folder, CLAUDE.md, README) by verifying every external claim (URLs, library names, quoted attributions, named principles, project-style rules) against authoritative sources and tightening prose one small pass at a time. Use whenever the user says "auto-improve X", "polish X over time", "loop improving X", "verify and
The exact command may vary by repository. Check the README on GitHub.
For the skill author
Drop this on your repo README
Shows your skill is listed on Skillteca, generates a backlink and trackable traffic.
[](https://www.skillteca.com.br/skills/auto-improve?utm_source=badge&utm_medium=readme&utm_campaign=badge)
One short email with only the new Pesquisa e Web skills. 4 minutes of reading, no spam, unsubscribe with one click.
You confirm your email on the first send. No spam. Unsubscribe with one click.
Auto-improve
Incrementally improve a target through a configurable loop. Each pass touches one file, makes a small verified change, and either stages it or records it as a checkbox finding. Many small verified improvements beat one big rewrite.
First thing every invocation: state check
Read .auto-improve.state at repo root.
Missing → this is the start. Run "Setup" below, then run one pass.
Exists → you're inside a loop iteration. Skip setup. Read config from state and
[Description truncada. Veja o README completo no GitHub.]
