chrisvoncsefalvaychrisvoncsefalvayCommunityยท2 Agent Skills Included

autostar

Iterative experiment loops that measurably improve any artifact

Runs structured experiment loops that iteratively improve code, prompts, documents, and configs against measurable quality tracks. Combines hard checks like linters and test suites with rubric-based scoring, so quality goals no longer require manual trial and error. Learns from every attempt through persistent memory, enforces budgets, and delivers a full report of what worked and why.
npx skills add chrisvoncsefalvay/autostar --all -g -y

All Skills in This Repository (2)

Pure Emerald Level Indicators

Frequently Asked Questions

FAQPage Schema
How to install autostar?โ–ผ

Run `npx skills add chrisvoncsefalvay/autostar --all -g -y` in your terminal to install the skill globally, then invoke it with /skill autostar in Claude Code.

What does autostar actually do?โ–ผ

It turns any measurable goal into a structured optimisation loop: it mutates your artifact, scores each attempt across multiple quality tracks, keeps improvements, and reverts failures until your budget runs out or the goal is met.

What kinds of artifacts can autostar improve?โ–ผ

Anything you can measure, including code quality, documentation, prompts, writing style, API design, and accessibility, using verifiers like pytest, pyright, eslint, lighthouse, or rubric-based judges.

Does autostar work outside Claude Code?โ–ผ

Yes. It ships runtime adapters for Claude Code, Codex, Gemini CLI, and Pi with full support, plus a reduced-support package for Claude.ai custom Skills.

Do I need machine learning expertise to use autostar?โ–ผ

No. You describe what good looks like in plain language, approve the plan and budget, and the loop handles experiments, scoring, and learning automatically.

Related Repositories in Software Engineering

View All in Software Engineeringโ†’