AI Skill Report Card
Converting Taxonomies to 3D Tech Packs
Quick Start13 / 15
Given raw taxonomy text like:
Fashion Design · 24
- Pattern Drafting
- Seam Allowance Calculation
- Professionalism
- Multi-tasking
- Fabric Draping
Produce:
- Parsed schema:
Fashion Design (parent, count=24)→ children tagged as[task],[activity], or[noise] - Filtered list: strip
Professionalism,Multi-tasking(generic buzzwords, zero technical utility) - Synthesis mapping:
Pattern Drafting→ maps to UV layout coordinates;Seam Allowance Calculation→ maps to mesh edge-loop split lines - Routed output: spatial tasks → Track A (3D mesh manifest); documentation tasks → Track B (printable worksheet)
No preamble, no conversational filler — output goes straight into structured manifests or vector layout specs.
Recommendation▾
Add a second full worked example for Track A output showing the complete mesh manifest text, not just a description of what it would contain — Example 2 is abstract rather than concrete.
Workflow13 / 15
Progress:
- Step 1: Lexical ingestion — scrub text, locate
·divider, isolate parent category from count metric - Step 2: Tag every child line as
[task](physical/executable),[activity](overarching workflow), or[concept](foundational knowledge) - Step 3: Run noise filter — remove soft-skill buzzwords (Professionalism, Communication, Teamwork, Multi-tasking, Time Management, etc.)
- Step 4: Synthesis matrix — pair each surviving 2D physical-craft task with its 3D digital equivalent (seam ↔ UV split, fabric grain ↔ texture flow, dart ↔ mesh deformation zone)
- Step 5: Route each synthesized node to Track A (spatial/3D) or Track B (systems/print) based on whether it produces geometry or documentation
- Step 6: Generate final asset — either a 3D mesh spec manifest or a monochrome print worksheet
- Step 7: Run validation checklist before marking output as export-ready
Recommendation▾
Clarify edge cases: what happens when a taxonomy entry is ambiguous between task/activity/concept, or when no children survive the noise filter (empty Track A/B output)?
Noise Filter Blocklist (Step 3 — reject these unless paired with a concrete technical verb)
Reject on sight: Professionalism, Multi-tasking, Communication Skills, Teamwork, Time Management, Adaptability, Leadership, Attention to Detail (standalone), Problem-Solving (standalone), Customer Service (unless tied to a specific deliverable).
Keep: anything with a measurable action + object (e.g., "Calculate seam allowance", "Audit vertex penetration", "Optimize polygon distribution").
Synthesis Pairing Reference (2D craft → 3D digital)
| Physical/2D Concept | Digital/3D Equivalent |
|---|---|
| Seam allowance | Mesh edge-loop offset / split line |
| Dart placement | Deformation weight zone |
| Fabric grain direction | UV texture flow alignment |
| Pattern piece boundary | UV island border |
| Fit ease (ease allowance) | Collision clearance threshold (mm) |
| Garment size grading (XS–XXL) | Blend shape / scale factor targets |
| Flat pattern layout | UV unwrap layout |
| Stitch line | Quad topology edge loop |
Output Format Rules
Track A (3D Mesh Manifest):
- Units: 1 unit = 1mm, absolute metric
- Size range always: XS, S, M, L, XL, XXL
- Must specify: collision clearance threshold, UV stretch tolerance (<5%), polygon budget, export formats (.OBJ, .FBX)
- Quad topology only at joints/deformation zones — no stray triangles on bend paths
Track B (Printable Worksheet):
- Canvas: locked 8.5×11", monochrome only (#000000 on #FFFFFF)
- No gradients, shadows, textures, or faux-3D — flat vector only
- 4-point vertical spacing grid for all typography/layout blocks
- Text converted to outline curves before final export
- Always include a two-column "Manual Human Task" vs "Automated Engine Scope" split table
- Always include a final status block:
[ ] CONFIRMED PASS [ ] REJECTED — RE-RUN PIPELINE
Examples14 / 20
Example 1:
Input: "Costume Design · 18: Pattern Grading, Seam Finishing, Professionalism, Cloth Simulation Setup"
Output:
PARENT: Costume Design (n=18)
[task] Pattern Grading → maps to → Size Grade Scale Factor (XS–XXL blend targets)
[task] Seam Finishing → maps to → Edge Loop Weld Verification
[noise] Professionalism → REJECTED (no technical utility)
[task] Cloth Simulation Setup → maps to → Track A: Draping Simulation Loop Config
ROUTING: 2 nodes → Track A (spatial), 0 nodes → Track B, 1 rejected
Example 2: Input: Request for entry-tier printable worksheet on mesh clearance checks Output: Full ASCII-style monochrome worksheet with PROJECT IDENTIFIER header, numbered checklist steps (Load Bounding Volumes → Enforce Clearance → Verify Seam Topology), Human/AI boundary matrix table, and status sign-off block — formatted per Track B rules above.
Recommendation▾
The validation checklist mentioned in Step 7 of the Workflow is never actually defined — add the concrete validation criteria/checklist referenced there.
Best Practices
- Always resolve ambiguous target software (Blender vs. Marvelous Designer vs. Unreal) before writing automation steps — ask once, then commit to that toolchain for the whole manifest.
- Keep parent-child count verification strict: if child task count doesn't match the stated metric (e.g., "·24"), flag the discrepancy before synthesis.
- Default to bundling: pair a free low-effort worksheet with a premium high-effort 3D asset pack when discussing product strategy.
- When in doubt about a skill's utility, ask "does this produce a measurable geometric or documentation output?" — if no, cut it.
Common Pitfalls
- Don't let decorative styling creep into Track B outputs (no drop shadows, no paper texture, no skewed backgrounds) — monochrome flat vector only.
- Don't leave soft-skill buzzwords in the synthesis matrix "just in case" — they dilute the manifest's technical density.
- Don't generate 3D specs without stating the clearance threshold and UV stretch tolerance — incomplete specs fail validation.
- Don't merge Track A and Track B outputs into one file — they serve different production pipelines and export formats.