Start with the job to be done

The working assignment is to direct a coherent three-shot paper-boat sequence for a short lesson opener for viewers watching a clearly labeled generated concept clip. That sentence is narrower than “use AI video generation.” It identifies a deliverable and a reviewer. Write a definition of done with three layers: the output must satisfy the audience's need; factual or functional claims must be traceable; and a named person must own the final decision. Generate short shots separately and edit them together. Consent, copyright, trademark, music rights, and clear labeling are production requirements—not cleanup after export.

Start by separating tasks. The model may draft, classify, transform, compare, or suggest. It may not silently approve, publish, grade, deploy, cite, or consent on someone's behalf. For this assignment the authoritative material is shot duration, aspect ratio, subject bible, action, camera, light, continuity anchors, negative constraints, and audio plan. Anything absent from those inputs is either an explicit assumption or an unanswered question.

Use this prompt as a realistic starting contract:

Create a 6-second 16:9 shot: red paper boat in a shallow rain puddle; it drifts past one yellow leaf; low close-up tracking left to right; overcast realistic light; calm continuous motion; no cuts. Preserve boat color and fold shape. Avoid people, text, logos, camera shake, and copyrighted characters.

Notice what the prompt does: it states the setting, limits the output, names forbidden behavior, and requests evidence that can be reviewed. It does not ask the model to “make it amazing.” If a constraint matters, make it testable. Replace “be accurate” with a source boundary, formula check, test command, rights ledger, or approval step.

A useful response would look like this: A bounded shot card with one action, one camera move, stable subject anchors, and negatives that can be checked frame by frame. That description is intentionally observable. “Looks good” is not acceptance. The operator must review first/middle/last frames, check anatomy and physics, compare continuity anchors across shots, inspect logos/text, and verify disclosure and asset licenses. Keep the source material beside the draft so review means comparison, not memory.

Run the work through brief → storyboard → generate shots → continuity pass → edit → disclose. Stop when an authoritative input is missing, a high-risk claim lacks evidence, private material cannot be safely removed, or the proposed action exceeds the permission granted. Escalation is successful workflow behavior, not model failure.

Common framing mistakes are too many actions per shot; character drift; warped text; impossible motion; undisclosed synthetic footage; unlicensed music or style imitation. Prevent them by writing a one-paragraph job card: user, decision, deliverable, source of truth, constraints, reviewer, and stop condition. This card becomes the anchor for every later prompt.