Shortfilm Desk

Somebody already made this, and wrote down what it cost. The upstream library ships twenty-six genre templates and six worked examples, and the examples record something almost nobody publishes: how many rerolls each pattern actually took before it produced a usable clip - two or three at the low end and twenty-plus at the high end, for the same shot. That variance is the budget, and knowing it before you start is worth more than any amount of prompt polish. So the free in-page matcher ranks the whole library against your idea with the arithmetic shown: which tags matched, what each was worth, and what the nearest recorded pattern cost. A template with no tag hit never appears at all - length and shape are tie-breakers between plausible answers, not reasons to be one. Then the seven hard rules run on whatever you write: a real camera and lens named, the breathing sentence verbatim rather than shortened to 'handheld shot', the sound line, at least two imperfections because too perfect reads as CG, none of the thirteen praise words a model cannot form an image from, no effects pile-up at the close, and no named property - substituted rather than deleted, so the aesthetic the name carried survives. It also knows what each of eight models will filter, how long it will run, and where its negative prompt goes, which has three different answers rather than two. Five lanes: find the template, write the five-stage prompt, expand the storyboard, rewrite it for the model's filter, and audit one you already have. Lanes derived from the open-source MIT agent skill jnMetaCode/ai-shortfilm-prompts. The reroll figures are that library's own notes, not measurements taken here. Not affiliated with or endorsed by jnMetaCode, Mx-Shell, or any property named in a substitution table.

Every public app is built from a security-scanned skill and must pass a clean scan — skill and frontend — before it can be listed. Have a skill of your own? Turn it into an app — or read the step-by-step walkthrough.