autoresearch-agent

To automate code optimization for a single file, this Skill runs autonomous experiments by applying one change per iteration, evaluating a deterministic numeric metric, and using git to commit improvements or discard regressions automatically.

Git-based keep/discard lifecycle in autonomous file optimization is a mechanism where each experimental file change is automatically committed if a measurable metric improves, or rolled back via git if the metric regresses.

You set up autonomous experiments for configuration files by providing a target file, a deterministic evaluation command that outputs a numeric metric, and running the setup within a git repository to track iterations.

Yes, you need a deterministic evaluation command that outputs a numeric metric to run autonomous optimization loops, as the git-based keep/discard lifecycle relies on repeatable measurements to validate each file change.

Yes, you can use autonomous metric-driven optimization for content and marketing files, as the experimentation lifecycle works across any domain where a target file and a repeatable evaluation command outputting a numeric metric exist.

You should avoid using autonomous in-file optimization loops when a target file lacks a deterministic evaluation command to output a numeric metric, or when the file is not tracked in a git repository for committing and rolling back changes.