Free Graph Ops upgrade for safer AI-generated repairs

I built Code Factory for my own needs after too many assistant-generated changes passed tests and then failed when I actually used them. The wasted time and frustration pushed me to put planning, evidence, and review into one proof-first workflow. I am sharing it free.

v0.32 adds a Counterfactual Arena to Graph Ops. It compares 2–12 supplied repair candidates, rejects failed proof, surviving mutants, scope escape, or changed receipts, and explains exactly why the smallest eligible repair won. It cannot generate or apply code; the human remains in control.

Solo builders get a readable choice before accepting an AI fix. Teams get a reproducible, hash-bound decision record. Time, token, and cost savings are shown only with an exact paired baseline; otherwise they remain “Not measured.”

I also used the open Prestige skill on the UI. In my own observation it materially improved the visual hierarchy and readability; the repo records its deterministic design checks without presenting that observation as measured conversion.

Screenshot: https://raw.githubusercontent.com/zrk222/code-factory/v0.32.0/docs/assets/marketplace/graph-ops-proofsearch-controls.png

Release: https://github.com/zrk222/code-factory/releases/tag/v0.32.0