# Specwatch release film storyboard — unpublished candidate

Scope: a real Specwatch CLI run against the disposable `fixture/src/math.ts` and an existing hand-written `math.test.ts`. The local Responses-API stub is deterministic and **not Ollama or a real model**. The film may demonstrate file behavior and CLI output; it cannot assert test quality, model quality, or coverage improvement.

Working release headline: **Preview before generation**.

| Beat | Product frame | On-screen annotation | Evidence gate |
| --- | --- | --- | --- |
| 1 · Source | Real terminal showing source file and existing test hashes | `Synthetic fixture · existing hand-written test` | Retain fixture files and SHA-256 |
| 2 · Preview | `specwatch gen src/math.ts --dry-run --no-validate` output | `Preview path; original test bytes unchanged` | Same test SHA-256 before/after; record exit code |
| 3 · Write | Normal `gen` output and a readable before/after test diff | `Normal generation can replace an existing test` | Exact CLI output and diff; no quality claim |
| 4 · Watch | A deliberate source edit followed by watch regeneration | `File change triggers another draft` | Timestamped source edit and watcher output |
| 5 · Boundary | Coverage/CI command against an absent report | `Missing coverage report is not a fail-closed gate yet` | Actual command and exit status |

Paired light/dark film and stills must show the same real terminal outputs at legible scale, with only restrained camera moves and annotations. Poster should frame the actual dry-run line and its unchanged hash proof, use a release-specific headline, and carry a prominent `SYNTHETIC FIXTURE · SIMULATED MODEL` label. Narration is pending an approved Cartesia key/voice. Do not publish or describe this as finished media before run, review, render, and caption QA.