Scribble Works marketplace taxonomy + site tree — proposal

Why this is in the vault

Founder rulings by iMessage 2026-08-31 ~15:24 ET reframed the catalog: no predefined packs — a marketplace of individual pages/games where the parent selects 6 to build their own pack, with search and filtering. He asked for a taxonomy proposal to lock in early, warning that "filtering categories can turn into ai slop without guardrails." This note is that proposal plus the site tree he asked about (build tree → stub pages → fill screens iteratively — he proposed the approach, Ray endorsed it). Status is proposal until he locks it; nothing here is a founder ruling except where marked.

Founder rulings captured (2026-08-31 15:24, iMessage — these ARE rulings)

1. Marketplace of individual pages, parent picks 6 to form a pack. No predefined packs.
2. Search + filtering required; taxonomy must be locked early with guardrails against category slop.
3. Ray is overused on the current site — "same sticker stuck all over." Needs situational, creatively varied poses.
4. Copy must not market toddler-only. The 3-4yo content is where we started (his daughter), not the audience ceiling.
5. Build a site tree, stub the pages, fill screens in iteratively (his proposal, confirmed).
6. Cream background: not in love with it — needs alternatives.
7. Wordmark v2: "Works" larger/heavier so the two words carry equal weight; "Scribble" currently reads sharpie-on-whiteboard — should be messier, waxier, crayon-like.
8. Overall: v2 is "a nice improvement over v1."

Marketplace UX anatomy (founder, 15:28 iMessage follow-up — rulings)

• Listing page: all worksheet tiles, searchable, likely paginated.
• Detail page per worksheet: "Add to workbook" + "Customize" — the generative-AI hook (regenerate the worksheet personalized). His words: "our generative ai hook." This is the differentiator and needs its own design + product round.
• Pull-out tray (right side) showing the in-progress collection while browsing.
• Accounts build a collection per day → "a whole content calendar." This partially answers storyboard Q5 (is-the-plan-a-screen): the plan is a calendar of workbooks. Accounts become "my kid's week," not order history.
• Founder 15:55 follow-up: 6 games/day is his lived baseline with his daughter ("about right for keeping attention but providing some variety"); open to a better number or parent-decides. Wants a hard number so unselected slots render as outlined placeholders, plus a magic-wand button that auto-finishes the remaining slots ("finishes out the curriculum for them without the parent having to hunt the whole marketplace"). Wants to steer clear of worksheet/workbook — fun-first, "not so much feel like learning even though they are."
• Ray recommendation (sent 15:57, awaiting sign-off): hard 6 (his lived data + the slot mechanic and wand need a fixed frame; flex by age band later if ever, not by preference). Terminology: one page = game, daily set = playset (compresses his "playground set," toy-aisle word, pairs with game, cashes in the "paper playground" marketing line). Rejected: activity, collection, pack, bundle, work-anything. Sign-off unlocks schema + tray design + wand spec.

Proposed taxonomy — 5 facets, closed vocabularies

The anti-slop design: a FIXED number of facets, each a CLOSED list. Generator agents pick from the list and can never invent values. The taxonomy lives as one versioned file (taxonomy.yaml in the scribble-works repo) that feeds BOTH the site filters and the content-generation prompts, so filters and content cannot drift apart.

Facet 1 — Age band (single select per page)

3-4 · 5-6 · 7-8 · 9-10

One primary band per page, optional one adjacent "stretch" band. Bands, not exact ages — exact ages invite endless micro-slicing. (Catalog age bands are a different surface from the parent-portal child profile, which per the [[2026-08-31-studio-charter|charter]] keys on birthday.)

Facet 2 — Skill (what it builds; 1 primary, ≤1 secondary)

Letters & Phonics · Reading & Words · Numbers & Counting · Math · Shapes & Patterns · Writing & Tracing · Science & Nature · Logic & Puzzles · Art & Drawing · Scissors & Motor Skills

The educational spine — this is the facet a curriculum-minded parent filters by first.

Facet 3 — Activity type (what the kid does; exactly 1)

Tracing · Coloring · Matching · Maze · Cut & Paste · Count & Tally · Word Game · Drawing Prompt · Game for 2+

Skill and Activity are deliberately separate axes: "Numbers via a maze" and "Numbers via tracing" are different pages to a parent whose kid hates tracing.

Facet 4 — Theme (the slop magnet; ≤2 per page, additions gated)

Starter set of 12: Animals · Dinosaurs · Space · Ocean · Vehicles · Fairy Tales & Magic · Food & Cooking · Seasons & Weather · Holidays · Sports · Music · Everyday Life

Theme is where AI-generated categories metastasize ("Enchanted Woodland Friends," "STEM Adventure Buddies"). Hard rule: agents may not add themes. A page that fits no theme gets Everyday Life or ships themeless.

Facet 5 — Practical flags (multi-select, fixed set)

Low ink · Crayons only · Needs scissors · Needs glue · 2 players

The kitchen-table facts a parent checks before printing. Maps to the ink-cost design constraint from the [[2026-08-31-brand-interview-r1|brand interview]].

Guardrails (the actual anti-slop mechanism)

• Facet count is frozen at 5. A sixth facet requires a founder decision, not an agent judgment call.
• Closed vocab, enforced in code: the site build validates every page's frontmatter against taxonomy.yaml and fails the build on any unknown value. Slop cannot ship silently.
• Cardinality caps: 1 age (+1 stretch) · 1 skill (+1 secondary) · 1 activity · ≤2 themes · any flags.
• Additions are demand-driven and batched: a candidate value goes to a review queue; it graduates only when ≥5 real pages want it AND the founder approves the batch. No single page ever justifies a new category.
• One source of truth: filters, page frontmatter, and generation prompts all read the same versioned file. Renames are migrations, not edits.

Proposed site tree (stub all, fill iteratively)

scribbleworks.co
├── Home                    (parents lead; kid energy shows in the work, not the chrome)
├── Browse                  (the marketplace: filter rail = 5 facets, search box, paper-grid)
│   └── /page/[slug]        (page detail: framed preview, facet tags, "Add to pack")
├── Pack builder            (persistent tray: pick 6 → name it → download PDF)
├── How it works            (print-at-home pitch, 3 steps, sample page free)
├── Our story               (the daughter origin, Grammy, the why; real-photo surface)
├── Portal (stub)           (account · kids · past packs — design gated on storyboard answers)
└── Help / FAQ              (printing, ink, licenses, refunds)

Build order: tree + stubs first so navigation is real and every screen has an address, then fill screens one design round at a time (Browse first — it carries the taxonomy).

Resolution log

• 15:57 founder: TAXONOMY LOCKED — "I'm good with your proposed taxonomy. We may add to it." Additions flow through the review gate (≥5 pages + founder batch approval). He singled out the practical filters as the facet he really likes.
• 15:57 founder: materials → add-ons idea — extra equipment a game needs (e.g., furry pipe cleaners) becomes purchasable add-ons on the site for parents who don't want to think about sourcing. Ray's structuring (sent 15:59): upgrade Facet 5 from flags to a structured per-game materials list; one source powers the filter, a "what you'll need" strip on the detail page, and the add-on upsell. Commerce lane is post-launch: affiliate links first (zero inventory, validates demand), stocked only if clicks earn it. Data structured now, nothing built.
• 16:07 founder: TERMINOLOGY LOCKED — one sheet = game, collection of six = playset. New ruling in the same message: every playset also gets a title page with a parent-facing writeup of what the playset offers — "lucky 7" printed pages per playset (6 games + 1 title page). This gives the queued parent-overview template its real identity: the playset title page (parent writeup, the 6 games listed, materials-needed pulled from taxonomy, printed as page 1).
• 16:07 founder: affiliate add-on approach APPROVED as the low-risk lane, agreed as a follow-on after initial buildout — not in scope now.

Open questions for the founder (in the lock-in message)

1. Pack size: locked at exactly 6, or "up to 6"? (Pricing implications either way.)
2. Age bands: does 3-4 / 5-6 / 7-8 / 9-10 match how he thinks about the ladder, or prefer school-stage labels (Preschool / Pre-K / K / 1st-2nd...)? US-school labels read regional; number bands travel.
3. Any facet or value he'd cut? Smaller is safer to lock.