Use When
- A prompt uses benchmark videos, contact sheets, frames, product images, persona images, style boards, or audio references.
- The model needs explicit learn/do-not-copy boundaries before generation.
- Some references should bind identity while others should be inspiration-only or excluded from the first test.
Do Not Use When
- Do not decode the hook or opening mechanism here; use
reference-decode. - Do not force every reference into "learn from".
- Do not treat identity references as loose style inspiration unless the user says so.
Core Rule
Every reference set must answer:
- Which references bind identity or consistency?
- Which references are inspiration-only and must not be copied?
- Which references are intentionally excluded for this test?
Default assumption: persona, product, and audio references are strong bindings unless the user explicitly marks them loose, benchmark-only, or weak guidance.
Default Workflow
- Inventory the reference set: hook clip, hook-first contact sheet, style board, product image, persona image, audio reference, or other source.
- Decide the test purpose: hook rhythm, camera realism, product proof, persona continuity, or another narrow target.
- Classify each reference as
binding,inspiration-only, orexcluded for now. - Create a compact machine-readable contract when it is useful.
- Print or return the contract before final prompt writing.
Output Shape
The artifact contains testPurpose, mayLearn, mustNotCopy, and
excludedReferences. Add user-specific binding notes around that output when
the prompt has real product, persona, or audio references.
Stop Conditions
- Stop when required user intent, source evidence, or owned input artifacts are missing and guessing would change the result.
- If an owned CLI or script command fails, report the exact error and stop. Do not bypass the failure with metadata-only answers, readiness probing, local payload rewrites, fallback providers, or unpublished tools.
Handoff
- Hand
contract.jsontovideo-request-architector the prompt writer. - If the reference's hook logic is still unclear, hand to
reference-decodefirst.
Public Command Boundary
- Choose the smallest matching command or workflow from the user input and run it directly.
- If an owned CLI or script command fails, report the exact error and stop. Do not bypass the failure with metadata-only answers, readiness probing, local payload rewrites, fallback providers, or unpublished tools.
- This public skill is instruction-driven. Produce the artifact described by the workflow directly from the available evidence.
- Do not call private provider/runtime paths or unpublished local tools.
- If the CLI returns a quote-confirmation challenge, run
postplus quote confirm --json --challenge-file <challenge.json>and retry with the returned token.