aeon-autoresearch

Automate AI skill improvement by generating four parallel variations, scoring them against a weighted rubric, and applying the highest-scoring version to enhance signal-to-noise ratio without manual rewriting.

Parallel variation scoring for skill evolution creates four distinct approaches—better inputs, sharper output, more robust logic, and a fundamentally different method—then scores each to safely apply only the winning variation.

Update AI skills hitting deprecated APIs automatically by running an evolution command that generates parallel variations, targets the deprecated endpoints, and applies the best-scoring functional replacement.

You can refresh stale AI skills without downgrading core functionality because the evolution process includes a safety guarantee that cleanly aborts if no generated variation scores higher than the original.

Automated skill evolution works for robustness enhancements by explicitly generating a parallel variation focused on making the skill more robust, then scoring it alongside alternatives to select the best overall improvement.