LOADOUT / Hands-on reports
Real tools, actual engine captures and a clear verdict. Find out what worked, what needed a fix, and where the evidence stops.
Latest reports / newest first
An agent skill for Three.js and Blender scenes. We built three small playable games with it and you can play all three here.
The catch: it made the games more correct, not faster. Three builders scored it 2, 2.5 and 2 out of 5. On Windows its scripts need python, not python3.
Read the test
734 animated pieces drawn in letters, plus a small kit for drawing your own. 733 passed the built-in check. Six ran at once from our own site.
The catch: its copy-a-prompt feature loads the script from the maker's site. Self-host it instead. It is an art library, not a game engine.
Read the test
A useful first lookup for low-poly props and shape variants. Four imports matched their geometry metadata.
The catch: the original materials rendered white. A persistent import correction restores colors. Tested color-only exports kept default paint.
Read the test
Pixel Art Lighter kept the painted crate face flatter. Laigter remains useful for command-line batches.
The limit: one 48 × 48 sprite, default settings, 13 light positions and normal maps only. No whole-tool ranking.
Read the testA Field Test means we actually ran the named tools in the setup shown. It does not mean every feature, asset or device passed. Major caveats stay beside the verdict; each full report identifies the tested versions, hardware and remaining checks.
Documentation-only comparisons are labeled separately. The original CROSSCHECK verdicts came from evidence reviews and remain their own department.