vc:autoresearch

Automate metric optimization by configuring an autonomous iterative loop that runs repeated experiments, evaluates a mechanical score, and logs results. The loop automatically rolls back Git commits if a regression is detected, ensuring measurable metrics like test coverage or bundle size improve traceably.

An autonomous optimization loop iteratively applies metric experiments and uses Git to back experiment history. It creates commits for each trial, evaluates the metric score, and automatically rolls back the Git history if a regression occurs, providing traceable lineage for reproducible research.

Yes, you need a clean Git working tree to run automated metric experiments. The loop requires a Git repository to back experiment history and execute rollbacks on regression, along with a SKILL.md frontmatter defining the Scope and a Verify command outputting a single numeric metric.

Set up a Verify command by defining a shell command in your SKILL.md that outputs a single numeric metric. The autonomous loop executes this command to judge each iteration, applying changes only if the metric improves and logging the outcomes to loop-results.tsv.

Yes, you can limit modifications by defining a Scope of editable files in your SKILL.md frontmatter. The autonomous optimization loop restricts its iterative experiments to this specified scope, preventing unintended changes to your repository while optimizing the target metric.

If an iteration causes a metric regression, the loop automatically executes a Git rollback to revert the changes. This ensures the mechanical score never degrades, maintaining traceable experiment lineage and preserving the last known optimal state in your repository.