Autoresearch

Automate skill evolution by generating four distinct variations of a target skill, scoring each against a rubric, selecting the winner, and committing the updated version as a pull request. This removes manual iteration from version control updates.

Rubric evaluation for pull request generation works by scoring four generated skill variations against predefined criteria, selecting the highest-scoring winner, and committing only that version. A rationale trail is preserved for audit.

You can use automated variation generation for any target skill that has a definition file stored at the designated path. Simply set the target skill name to begin the research and evaluation process.

The best way to iteratively improve skills without manual review is to automatically generate multiple variations, evaluate them against a scoring rubric, and merge the winning version via PR. The system logs outcomes to memory for tracking.

Automated skill improvement preserves an audit trail by logging results to memory and maintaining a rationale trail. This ensures every pull request commit has a recorded justification for the selected winning variation.

You need automated rubric scoring for skill variation proposals to objectively evaluate multiple competing theses, select the highest-quality version, and ensure only validated improvements are committed to version control.