Client intake — docade

Stages 1–2. Drafted 2026-08-17 from docade's own documentation, not from an interview: assets/manifest.json, assets/reference-sheet.txt, and docs/14-ASSET-MANIFEST.md. Sources are cited so every claim can be re-checked against the client's repo rather than trusted.

Marked ⬜ where a decision is genuinely open. Those are the only things that need a human.

Identity

Request source

docs/14-ASSET-MANIFEST.md §6 maps paths to database columns: profiles.avatar_key, tasks.icon_key, prizes.image_key, collectibles.image_key. Keys are kebab-case and stable forever; renaming one is a migration, not a file rename.

Delivery rule for this client: WootBuild never invents a filename, and art deliver must refuse any path that is not a declared manifest key. Art delivered to an unknown path renders broken in production and nothing catches it — docade's own §6, and the failure their check-assets script exists to prevent.

Stage 1 — What the assets do

Function map

Sources: reference-sheet pieces 1–13; 14-ASSET-MANIFEST.md §3.5 (80px), §3.8 + lib/prizes/art.ts (gradient grounds), 04-LYLE-COPILOT.md §2 (register).

Hard constraints

• Filenames are database foreign keys. Above.
• 80px is the judging size for collectibles, not 512. Check before approving.
• Prize art is judged on its committed gradient, e.g. #7c5cff → #4ac8ff.
• Silhouettes must not be commissioned. §5: 18 uncollected-state images are a deterministic alpha-channel transform (flatten to one fill at low opacity). One script, zero art budget, and they can never drift from the source art. This is WootBuild's job via sharp, not a generation job.
• Reduced motion is replace, never remove (§2.5 of their motion grammar).
• Motion direction already carries meaning: up+fade = value leaving, down = arriving, scale = now most important. New motion speaks that grammar or is noise.

Stage 2 — Volume and staffing

Volume profile

⚠ Launch counts are sculpts, not entries. Q25 made each colorway its own collection entry, so shipped entry count is unknown until Q44 lands. A manifest counting only launch "reads as complete on the day the library starts growing."

Lane grouping — 19 delivery categories → 10 production lanes

Lanes group by how a thing is made, not where it lands.

Split from the first pass after reading the reference sheet: prize left plush because it is the only register on a saturated ground, and crane left ui-art because it is motion. Both splits came from the client's own doc, which is the method working.

Best-artifact analysis, and the staffing gaps

Decided with the roster ignored, then diffed.

On gap 1 — quoting docade directly, because it changes our sequencing:

"There is currently no way to actually LOOK at a candidate piece in place… A style is judged at the size and against the background it will ship at — a stuffy at 80px on the collection grid, a task icon in the Today list beside real copy, a prize on the wall next to its ticket price… A harness that drops candidate files into the real surfaces at real sizes is a prerequisite for Step 1, not a convenience."

WootBuild's review UX is a contact sheet — a grid at uniform size. That is exactly what docade says is insufficient. This is a studio capability gap, and it independently confirms the Stage 1 "sits beside" principle.

On gap 3 — motion. 12 pieces are built as hand-written CSS in globals.css and are good: documented, consistent, reduced-motion honoured. 9 are unbuilt. Framer Motion, Rive and Lottie are all named across their docs and none is installed. Their Q45 narrows the real question to whether the crane and rarity reveals need a runtime state machine — and is explicitly blocked: "Do not answer it before piece 2 of the reference sheet comes back."

The critical path

Reference sheet piece 2 — plush octopus, ×2 colorways — is the highest-leverage single asset in this engagement. It settles:

• Q44 — whether a recolor is a fraction of a sculpt's cost, i.e. whether colorways-as-entries works and 150/year is doable. "If it is a repaint, the entire growth model costs a multiple of what was assumed."
• Q45 — the motion library decision, which their doc couples to Q44.
• Our gap 2 — colour separability.

Better to learn it on piece 2 than at sculpt 18, which is their reasoning and it is correct.

Sequence: visual mission (Stage 3) → placement harness (gap 1) → reference sheet with piece 2 first → volume.

Decisions — 2026-08-17

⬜ Still open — not ours to answer