A test is four things
Download the official build and unpack it. Start it, and get a window, a listening port or a working command output. Drive the feature the article is about, and watch what it produces on disk. Take screenshots of that, not of a mockup.
| Check | Requirement |
|---|---|
| Install | Official release, unpacked and started on the recorded OS |
| Core feature | Actually run, with the produced files or responses inspected |
| Screenshots | Captured from the running software, never rendered from HTML |
| Package | Portable zip with a SHA256 you can verify locally |
What is left out
Anything not run is marked as not verified in the writeup, including features that look obvious. A build that fails to reach its main window is reported as a failure, with the step it stalled on.
Two languages, one test
Some tools have both a Chinese and an English entry. They describe the same install and share the measured numbers; only the writing differs. Pages carry hreflang tags pointing at each other, so search engines see one test in two languages rather than two competing pages.
The Chinese board has the full archive.