Independent draft review — Contractor+ preflight and client questions

Date: 2026-09-10. Requested model/effort: Astra ultra. Observed exact runtime setting: unverified. Scope: written evidence and current drafts only; no original-media watching/listening, email send, media generation or spend.

Verdict: approve the client questions with one accuracy edit. No missing production-blocking question was found. The user has already authorized sending Justin the necessary questions; this review does not require another permission request.

Reviewed QUESTIONS.draft.txt (17 lines), PROJECT-PREFLIGHT.draft.md (38 lines), CLAIMS-AND-ASSETS.md, recovered history ending, actual stale five-customer script passages, nested core inventory and Jobber packet-match receipt.

1. Email opener, line 3 — correct the tool-execution claim. “Checking the original media with PromptLab and Reverse Prompter” can imply both have already run. Unless execution is evidenced, use: “I've gone through the recovered history and package. We're reusing PromptLab's earlier Jobber teardown and checking the originals against lessons from our own video work before new generation.” The current Jobber receipt confirms equal encoded video/audio packets, with different containers; saying simply that both files are byte-identical would be inaccurate.
2. Question coverage is sufficient and avoids duplicated intake. The five groups cover first story/placements/review date, genuinely missing production layers/current UI/better customer footage, growth definition/evidence/reuse approval/story approver, current offer/CTA, and the pilot budget. Brand PNGs are already present, so asking only about changes is correct. ElevenLabs and quality preferences are carried forward rather than re-asked. Square and portrait are proposed without another format-selection burden. Follow-on story order can wait if the client needs to answer the pilot questions first; this is optional simplification, not a blocker.
3. Quarantine the old F1 copy in the project preflight. Add an explicit reference to the claim ledger's stale-script disposition. contractor-plus-5-customer-films.md:62 retains nineteen years; lines 74–76 describe a particular homeowner incident; line 93 presents a five-star review. The history reports the experience correction, and the source lane found no supplied-testimonial support for the incident/review. Treat these as an unapproved draft regardless of the “verified” label. The client email need not repeat all errors, but the new script must not reuse them.
4. Update the preflight's pending-evidence paragraph before final handoff. Originals and full history are now recovered; the missing items are the promised separate video production layers/current UI and remaining client decisions. Distinguish recovered inputs, completed deterministic extraction, direct media review, and unrun semantic/model work from each other. The draft appropriately labels later production as a sequence rather than a completed claim.
5. Keep the final publication boundary separate. The authorized action now is emailing the necessary project questions. The draft does not claim a film is complete, infer an old ten-film budget, or launch paid production. The pilot proposal is reviewable and respects the evidence limitations.

Strong elements retained: source-based customer facts, growth denominator/period, genuine customer footage, per-ratio type composition, narration fitted without speed compression, a representative proof before batch production, versioned outputs and final-file review. The real-footage recut/pickup-shoot alternative remains viable if existing assets cannot support the intended quality.

This review's scope is source/draft accuracy and question completeness. It does not certify current vendor capability, film quality, subjective voice quality, rights or final client acceptance.