davebcn87davebcn87Community·3 Agent Skills Included

pi-autoresearch

Autonomous experiment loops that optimize any measurable code metric

Runs autonomous optimization loops inside the pi coding agent: try an idea, benchmark it, keep improvements, revert regressions, repeat. Eliminates manual benchmarking, lost experiment history, and messy trial-and-error branches for targets like test speed, bundle size, build time, and model training. Persists every run to resumable session files and finalizes results into clean, reviewable git branches.
npx skills add davebcn87/pi-autoresearch --all -g -y

All Skills in This Repository (3)

Pure Emerald Level Indicators

Frequently Asked Questions

FAQPage Schema
How to install pi-autoresearch?

Run `npx skills add davebcn87/pi-autoresearch --all -g -y` in your terminal to install all skills in this suite globally.

What does pi-autoresearch do?

It lets the pi AI agent run autonomous optimization loops: it tries code changes, benchmarks them, keeps improvements, reverts regressions, and repeats until you stop it.

What can I optimize with pi-autoresearch?

Any measurable target works, such as unit test runtime, bundle size, build speed, Lighthouse scores, or LLM training loss.

Does pi-autoresearch survive restarts and context limits?

Yes. All progress is saved in a .auto folder with a prompt file and append-only log, so a fresh agent can resume exactly where the last session stopped.

How do I turn experiments into reviewable code?

Run the autoresearch-finalize skill, which groups kept experiments into independent git branches, each with metric improvements in the commit message.

Related Repositories in Software Engineering

View All in Software Engineering