Autoresearch Experiment Loop
Runs an autonomous try-measure-keep-or-revert loop to improve a measurable metric in your code.
- Repository
- github/awesome-copilot
- Stars
- 40,703
- Licence
- MIT
- Files
- 1 file
What it does
Once installed, the agent first interviews you to define a goal, a command that measures it, the files it may edit, and any limits. It then makes code changes one at a time, runs your measurement command, and keeps a change only if the metric improves, otherwise reverting it with git. Every attempt is logged to a results file so you can review what was tried.
Great for
- Developers who have a clear, repeatable benchmark, test count, or build time they want to optimize
- Anyone who wants to let an agent run many small experiments unattended on a git-tracked project
Skip it if
- One-shot tasks, simple bug fixes, or code review, which the skill itself says it should not be used for
- Projects without a measurable metric or without a git repository, since the loop depends on both and could waste time or make unwanted edits
Try asking
- “Use autoresearch to reduce the runtime of our test suite, measuring with the existing benchmark script”
- “Run an autonomous experiment loop to increase test coverage of the parser module, lower-is-better on uncovered lines”
- “Autoresearch: try to shrink the bundle size of this web app within the src folder without changing the public API”
Install in 3 steps
- 1
Download the skill
The ZIP contains the 1 original files, the MIT licence and an attribution note.
autoresearch.zip - 2
Add it to your agent
Claude Code (all projects)
mkdir -p ~/.claude/skills && unzip autoresearch.zip -d ~/.claude/skills/Claude Code / Agent SDK (this project only)
mkdir -p .claude/skills && unzip autoresearch.zip -d .claude/skills/Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.
- 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
- Define a precise metric command and output extraction method up front, and list out-of-scope files explicitly, since the agent asks for these before starting
- Set a maximum experiment count for your first run so you can review results before letting it run unlimited
- Review results.tsv and the git history afterwards; the loop commits each attempt and creates a new branch
What's inside
- SKILL.md
Redistributed unmodified from github/awesome-copilot 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.