Docs / 12
Honest limits and how Cinematic UI was tested
What Cinematic UI has not verified (Safari, Firefox, real phones, screen readers), where procedural 3D falls short, and how blind agent tests shaped it.
A skill that claims everything is a skill nobody can trust. These are the limits stated in the repository, plus how the skill was tested.
Honest limits
- Not verified on Safari, Firefox, real phones or screen readers. The kit and scripts were checked in headless Chrome. Smoothness cannot be judged headlessly.
- Procedural 3D struggles in close-up on familiar real objects. The skill asks for assets and keeps weak ones in wide shot, but it cannot conjure a product photograph.
- Gaussian splats are described, not built. The skill describes the technique but ships no implementation.
- Film mode renders silent unless you supply a track. Films are made with the companion skill cinematic-cut, which carries the renderer.
- It raises the floor and widens the range. It does not guarantee an award.
Because of the first limit, the delivery report is required to say what was screened and what was not, and the skill never describes a frame it did not look at.
How it was tested
Fresh agents with no other context were given the skill and a client brief. Their output is the showcase image in the README, unretouched: fictional briefs, real output. For 2.0 the briefs were a tea brand homepage, a twenty-second launch teaser rendered to video, an independent bookshop’s subscription page built with no 3D at all, and a copy-heavy consultancy page.
Their friction logs, an independent review by a second model, and a survey of other skills were folded back into the text and the kit before release. The kit and scripts are verified in headless Chrome with three 0.186, GSAP 3.15 and Lenis 1.3, covering:
- scene and cut logic
- frame-exact seeking in any order
- reduced motion
- keyboard order
- every look
- frame-by-frame video rendering
Later releases were tested the same way. App mode was developed from three blind tests (an analytics office, an order back office, an on-call console) that exposed the interchangeable-dashboard problem, then retested blind on the new rules.
The demos
The ten demo productions on this site are fictional products directed by fresh agents using only this skill, each with its treatment. Their brands, prices, customers and figures are invented and labelled as such on the pages. Their treatments also record their own remaining flags, for example dead-hold and flat-frame warnings that were judged by eye, and a grade-drift measurement that a six-colour palette cannot satisfy across six different paintings.