Turn Prompt Libraries into a reviewable shot contract.

Write the invariant shot intent first

For Prompt Libraries, state the subject, environment, action objective, emotional beat, and end condition in plain production language. These invariants keep general video prompts variants comparable instead of producing unrelated ideas. For this section 1, save a annotated source board; have the rights reviewer review action readability; record the failed case as well as the accepted one; and do not advance it beyond a reversible handoff until the named limitation is resolved. Keep libraries evidence separate from prompt assumptions.

Turn the topic into a repeatable method

Move from definition to a small practice sequence, then review the result against explicit craft, rights, and delivery checks. Keep vendor-specific behavior outside the method unless a source verifies it. For this section 2, save a frame review; have the brand reviewer review revision control; record the failed case as well as the accepted one; and do not advance it beyond a reproducible test until the named limitation is resolved. Keep libraries evidence separate from prompt assumptions.

Layer camera, timing, continuity, and sound

Add framing and camera movement only when they support the action. Mark temporal beats, preserve identity and environment anchors, describe sound as intent rather than a guaranteed feature, and name artifacts or changes to avoid. For this section 3, save a revision log; have the producer review source rights; record the failed case as well as the accepted one; and do not advance it beyond a reproducible test until the named limitation is resolved. Keep libraries evidence separate from prompt assumptions.

• Subject and action — record versioned handoff, product specialist, revision control, and approved master; keep libraries evidence separate from prompt assumptions.
• Framing and camera relationship — record asset ledger, fact checker, temporal stability, and approved master; keep libraries evidence separate from prompt assumptions.
• Beat order and end state — record decision memo, technical reviewer, visual hierarchy, and approved master; keep libraries evidence separate from prompt assumptions.
• Continuity, sound intent, and exclusions — record acceptance matrix, technical reviewer, delivery fit, and approved master; keep libraries evidence separate from prompt assumptions.

Create controlled prompt variants

Change one meaningful dimension per version, such as shot size, movement, pacing, or performance intensity. Label the hypothesis for each variant so review can identify which instruction caused the difference. For this section 4, save a prompt brief; have the motion designer review editability; record the failed case as well as the accepted one; and do not advance it beyond a reversible handoff until the named limitation is resolved. Keep libraries evidence separate from prompt assumptions.

Review the output without inventing model controls

Check whether the requested action reads clearly, the camera supports the beat, references remain coherent, and the ending is usable. A prompt template does not prove that any named model, duration, audio mode, or parameter is available. For this section 5, save a shot contract; have the brand reviewer review input fidelity; record the failed case as well as the accepted one; and do not advance it beyond a reversible handoff until the named limitation is resolved. Keep libraries evidence separate from prompt assumptions.

Keep the evidence ledger attached to the decision

Partial evidence was supplied, but it does not establish product support or a complete capability, customer, or performance claim. Record the source, verification date, claim scope, unresolved gap, and the decision that the evidence can support. Search demand must never be reused as capability proof. For this section 6, save a continuity sheet; have the legal reviewer review claim support; record the failed case as well as the accepted one; and do not advance it beyond a reversible handoff until the named limitation is resolved. Keep libraries evidence separate from prompt assumptions.

Build a specific test brief for prompt libraries

Start with a bounded text brief with visible constraints and an end state. Define one observable change, protected details, a stopping rule, and the named reviewer. The intended output is a reusable prompt contract with controlled variants. Test one variable per version, preserve the source and settings, and compare results at the actual delivery size instead of choosing from an unrecorded impression. For this topic test, save a asset ledger; have the art director review temporal stability; record the failed case as well as the accepted one; and do not advance it beyond a reproducible test until the named limitation is resolved. Keep libraries evidence separate from prompt assumptions.

• Primary query: prompt libraries; test record: prompt brief, brand reviewer, claim support, and editorial approval; keep libraries evidence separate from prompt assumptions.
• Editorial owner: keyword-expansion:0187; decision record: delivery checklist, motion designer, reference integrity, and approved master; keep libraries evidence separate from prompt assumptions.
• Source scope: a user-provided competitor-gap export dated 2026-08-05 supports topic prioritization only; source review: decision memo, post supervisor, motion coherence, and shot approval; keep libraries evidence separate from prompt assumptions.

Separate topic fit from product proof

A dated, user-provided competitor-gap export supports only the decision to cover “prompt libraries.” It does not prove audience demand, SEELE capability, third-party behavior, commercial value, or a likely outcome. Verify product-specific statements against current first-party documentation and a recorded representative test. For this source review, save a asset ledger; have the creative producer review delivery fit; record the failed case as well as the accepted one; and do not advance it beyond a stakeholder sign-off until the named limitation is resolved. Keep libraries evidence separate from prompt assumptions.