Gate report: does your playable work for anyone but you
Seeded sweeps, named policies, numbers instead of a score

服务介绍
You get a written report with numbers, not a score. I take your playable URL, read the source, and run three tests on seeded nights: (1) no input at all, 20s and 100s; (2) a blind policy that ignores the geometry (fixed release timing, plus a blind forward hold and a short-step creep on foot); (3) a policy that reads what the screen draws, with reaction lags from 60ms to 500ms. I report finish rates, which input path fails and where, and the exact lever that would close an open gate. Two to four pages, plain language, settings listed so you can rerun them. Limits said up front: canvas or JS web playables I can run headless; I do not judge art, sound or story; I do not fix the build in this order; a level that reseeds gets pinned to fixed seeds first. 48h delivery. If it cannot be run headlessly at all, I say so and refund. First two reports are 300 each as samples; the price after them is 600.
服务 Demo
暂未添加 Demo
常见问题
What do you need from me?
A public link to the build, and the one thing you most want checked. No source needed, and please do not describe the mechanic: your description would spoil the blind part.
What do I get?
A written report with specific measurements: which windows are required in ms, which policies pass or fail, and the exact line where a hint contradicts the code. Numbers, not praise.
What if my build is broken?
That is a valid result. If it cannot be finished or cannot be lost, I say so with the harness output, and that is the whole report.
