URL/Product to Explainer Video

Apply ../00-video-production-guardrails/SKILL.md first; its gates apply to every video produced from this workflow.

1. Input intake gate

Accept exactly one of two inputs: a single product URL, or a short product brief (written claims/features supplied directly by the requester). Do not proceed past this step without one of the two.
If given a URL, fetch and read the actual page content before any other step; do not proceed on the URL string alone or on background knowledge about the product or brand.
If given a brief instead of a URL, treat the brief itself as the primary source for step 2 — there is no separate page to fetch, but the same "only extracted/verified claims" rule in step 2 still applies to it.

Unlike a blog-to-video workflow, this workflow has no existing long-form source text to draw from — extracting the product's claims, features, and positioning from the fetched page (or supplied brief) is this workflow's own first step, done before any scriptwriting.
List every claim, feature, spec, and stated benefit found in the source, each traceable back to the specific page section or brief line it came from. Do not add a claim the source does not state, and do not soften or strengthen a claim beyond what the source says.
Where the source itself makes a claim that looks unverifiable or unsourced (e.g., "the #1 choice for professionals" with no citation on the page), carry that claim forward flagged as unverified rather than dropping it silently or treating it as confirmed. [FTC-AICLAIMS-01]
The output of this step is a claim list, not a script — do not blend extraction and copywriting into the same pass.

Draft the script and scene-by-scene breakdown using only claims from the step 2 claim list; a claim that did not survive extraction does not appear in the script. This is the same accuracy gate ../00-video-production-guardrails/SKILL.md applies to every workflow in this pack — a URL-to-video pipeline is most exposed to it here, since there is no pre-existing published post to serve as a faithfulness anchor.
For each scene, record what it needs to show or say and which extracted claim(s) it illustrates, so a later reviewer can trace every scene back to a source claim.
A claim flagged unverified in step 2 must clear an independent primary source before it can appear in the script; if it cannot be substantiated on request, exclude it from the script entirely. There is no hedged-inclusion option — ../00-video-production-guardrails/SKILL.md's content-accuracy gate does not treat a hedge as equivalent to substantiation. [FTC-AICLAIMS-01]