Reuse-trial status

One independent editing brief has a scoped pass with inherited source limitations; the second has a qualified reuse-success recommendation, not an unqualified source-fidelity or isolation pass. This is not 20 successful briefs or acceptance of the underlying source examples.

The first frozen-packet attempt selected a suitable recipe and authored a different-content input, but rendering failed because the delivered packet omitted project/remotion.config.ts. No movie or manifest was produced. The consumer did not repair shipped files or receive undocumented help. Its packet, command log and failed submission remain in packets/brief-01-attempt-01/; the protected-write watcher has stopped. brief-01-attempt-01.audit.json retains the failed mechanical audit. The packager now includes and binds every required runtime root file, with 27 passing focused regressions in ../evidence/packet-fix-20260905/. A fresh attempt must demonstrate real rendering; author-side tests do not turn the failed attempt into a success.

The fresh brief-01-attempt-02 packet gets past that packaging failure. Its consumer preserves every shipped byte and correctly rejects the exact word Sketch: measured207.3px against the source-sized row0 box of181px. Both attempted outputs remain absent; brief-01-attempt-02.audit.json records failure. This is an unsupported exact-copy case, not a successful edit. The catalog now exposes the actual text-box/font constraints before rendering; it does not enlarge the box or shrink type. The original brief and both failures remain unchanged.

brief-03-attempt-01 delivers the requested two-shot edit and two native helper renders. All three pass the mechanical audit, including complete embedded recipe equality and the stopped watcher's no-protected-write check. The final edit is 203 frames at 60/1, 1920×1080 and silent. Its fresh reviewer reports trialPass: true, with no new adaptation, timing, copy or continuity defect, including the 82→83 reset cut. Inherited paragraph glyph/sidebearing differences and underline source-review limits remain explicit. The review is ../evidence/reuse-review/brief-03-attempt-01/review.json and review.md. It does not accept those source differences or award completed-example credit. The earlier mechanical audit stays unchanged; its trialPassed: false records the then-pending visual check, not a rewritten later verdict.

Four brief files exist. Brief04 supplies shorter, different studio copy and the same permitted original pictograms. Its consumer selected the masked-name reel and rendered the exact copy and six supplied images: 300 native frames at 60000/1001, 1920×1080, silent. brief-04-attempt-01.audit.json passes the mechanical checks; the stopped watcher records no protected writes. Its fresh review retains the native choreography and closing cut, but measures inherited pill-span differences: 436/384px source/output at local15, 404/362px at59, 355/394px at118 and 384/352px at164. These are visible thresholded scanline spans, not recovered vector widths. They are not new consumer geometry edits and do not disappear because the renderer succeeds. The report is ../evidence/reuse-review/brief-04-attempt-01/report.json and REPORT.md. The consumer also discloses early npm commands with an unaudited default-cache setting; the watcher cannot certify writes outside the packet. This uncertainty remains separate from its verified unchanged shipped files. Retain this qualified result without granting unqualified fidelity, strict write isolation or completed-example credit. It does not rewrite the failed Morrow brief or create another counted source treatment. Author text-fit estimates are outside the consumer packet and are not trial results. Brief03 adds a two-shot, mixed-native-cadence editorial assembly; its authoring-time text-fit estimate is not an independent render or source-fidelity pass.

The exact copy in brief-02/BRIEF.md is now a known unsupported-input case for the paragraph recipe. Its loaded-font advances are approximately 3486px/3387px; the corrected recipe permits 3200–3460px/3120–3280px and requires a shorter second line. Do not quietly rewrite that brief or count a rejected edit as a successful trial. It can test the documented rejection behavior separately.

Successful trials must use a frozen delivered package, independent agent selection, unchanged shipped recipes/components, supplied original assets, and actual rendered output. Audit embedded recipes in multi-shot edit inputs as well as files on disk: an unchanged implementation with altered embedded tracks does not satisfy the reuse constraint. Preserve failed attempts and report any undocumented help; neither may be relabeled as an independent success.