
from auto-improve19
An autonomous, judge-gated loop that iteratively improves text artifacts using a rubric and a separate evaluator model to ensure verified gains.
Auto-Improve implements a GAN-like iterative loop for text optimization. Instead of simple rewriting, it mutates a file, grades the change against a specific rubric using a separate model, and only commits changes that are pairwise-judged as superior to the previous version.
Use this when you need a document (like a landing page, cold email, or prompt) to be measurably better according to a defined rubric, avoiding the 'slop' of generic AI rewrites.
improve.py which manages the mutation, scoring, and decision process.Designed for agents with shell access and Python capabilities, specifically those integrated with the Gemini API for evaluation.
This skill has not been reviewed by our automated audit pipeline yet.