FPV Route Desk

One-shot first-person FPV video prompts, checked as arithmetic rather than read as prose. Five lanes over one scene: choose the POV identity and lay out the stops, write the numbered-stop prompt, write the batch image-pack prompt, write a red-line world route, and audit a prompt you already have into a repaired one. The free in-page checker runs before you spend a credit and enforces the four things that have to agree: a timeline whose windows tile the whole shot with no gap and no overlap, numbered stops that ascend from 1 and match the declared target count, a POV identity that can physically perform every movement asked of it (a drone has no walking bob, a robot vacuum cannot climb stairs, a cat cannot float over a table), and an image pack whose count adds up to one first frame plus one reference per target plus an optional clean frame. It also reads an avoid list as an avoid list, so it never reports a prompt for the very thing the prompt forbids. Seven POV identities, two route modes with their own required negatives, and the documented target-count bands for 5-8s, 8-12s, 12-15s, 15-25s and 25s+. Lanes derived from the open-source MIT agent skill zhouwei713/fpv-immersive-video-prompting. Not affiliated with or endorsed by zhouwei713, Hermes Agent, ByteDance, Kuaishou, Runway or Google.

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.