Fashion Design — Systems, Capsules & Specs
The craft layer for apparel brand-building: wardrobe architecture, capsule
economics, garment spec language, and fashion-specific evidence discipline.
Brand-agnostic — bring your own brandDNA.
1. Wardrobe System Architecture
Structure the offer as 2–5 systems named for wearer states, not garment
categories or demographics (reference pattern: FORM/MOVE/FLOW/TRANSIT/REST —
studio to recovery, one daily arc). Rules:
- Every SKU belongs to exactly one system; system tag appears on site,
labels, and internal docs — this is the primary navigation axis
- Fit logic is a separate axis from garment type. Define 2–3 fit logics
(reference pattern: R = one shared pattern / T = same design, blocks A+B
where anatomy changes the pattern / C = body-specific essentials). The
naming is an ethical position, not a sizing chart — whichever
demographics-split convention exists in the category (men/women racks,
"for her/him"), the brand's own tests must forbid it
- Hero product logic: choose the least crowded, most describable garment
as entry wedge ("one trouser anyone can wear and describe to a friend"),
versioned like software rather than seasonally replaced
2. Capsule & SKU Math
- Capsule = designs × colors × size-count; shared patterns multiply cheapest
(one pattern serves all bodies; per-block patterns multiply by block count)
- Publish the arithmetic in the entity doc:
(shared designs × colors × sizes) + (block designs × blocks × colors × sizes) + … = total SKUs
- Set a hard design cap (reference: 10 designs first capsule). New styles
must retire or version an old one — SKU inflation is silent and fatal
- Margin guardrails: target GM band at RRP, wholesale floor, and a
no-launch-below floor; no-discount policy except end-of-line retirement
- Pricing position: anchor explicitly between named competitors, with the
reason each anchor exists
3. Garment Spec Language (tech-pack register)
Every garment spec states measurable claims, not adjectives:
| Weak |
Spec-grade |
| "ultra-light" |
92 gsm |
| "quick dry" |
overnight dry at 80% RH |
| "durable" |
40 wash cycles, no visible pill |
| "comfortable" |
air-conditioning transition tolerant (28°C Δ) |
- Fabric property checklist for tropical/humid markets: wick, dry-time,
crease, stretch-recover, AC-transition, wash resilience, pill/cling,
hand-feel (matte, not sport-flash)
- Label honestly: proposal-authored numbers are targets. Only
wear-test data upgrades them to claims. Keep a target/validated marker
next to every metric
- Construction language: invisible-reinforced seams, stress-point bars,
hardware tonal not branded; the signature detail is construction
(curved waistband, spine seam), never print
4. Technical Flats (drawing garments that match captions)
Draw flats as pattern-language, not fashion illustration:
- Off-white ground, ink outlines (1.5–2 units at 400×500)
- Silhouette-specific geometry: trousers (waistband + taper/flare path),
tees (drop shoulder + rib collar arc), outerwear (collar band + zip line
- pocket rectangles), knitwear (cuff ribs, drawcord), compression
(contour curves + gusset lines)
- Always include: grain arrow, 2–3 notches, interior type + drawcord where
real, seam-path of the brand's construction signature
- Spec line under each flat in mono: fabric + GSM + key property
(
165 GSM JERSEY · BONDED SEAMS · A/B) — data only, never slogans
- One Python function per garment type; caption-matching is enforced by
construction, not by hope
5. Fashion Evidence Discipline
- In-market validation beats desk research: capture real competitor
store reports (layout, rack split, materials, expansion count) as raw
notes; map each brand proposal claim → confirmed / contradicted /
not-evidenced, with differentiation per row
- Track competitor motion via daily intelligence (brand-narrative revamps,
textile-science waves) and route findings into the entity page's
competitive-watch section with dates
- Fabric partner sourcing is a Gate-0-class activity: source fabric before
design, not after
- Fit blocks require fit models per block and grading obsession — the
pattern exists only when a real body in real humidity proves it
6. Fashion Hiring & Application Spec
When hiring designers for the brand, the application is a design test:
- Exact-look submission count (forces editing skill — mirror the capsule cap)
- The task is to design inside the existing system (11th piece) with
palette locked, fit logics fixed, SKU math required
- A page arguing against the candidate's own design
- Rejection codes (numbered) for auto-disqualification: over-count,
demographic styling codes, off-brief, oversize SKU, logo-print reflexes,
no self-critique
- Scored rubric with pass gate + per-section floors
7. Production Partner Integration (on-demand supply chain model)
Modern indie fashion brands rarely own factories — they integrate an
on-demand sourcing/SCM partner. The reference model (verified from
Creabra's published terms, 2026-08) defines the pattern to evaluate any
partner against:
7.1 The capability chain to require
design brief → [optional fabric lab / heritage textiles] → material sourcing → sampling → production → QC/final inspection → shipping
7.2 The commercial structure to expect
- Tiered sample development with published fixed prices by complexity
tiers (reference: standard / structured / complex; specialty fabrics &
haute couture quoted separately)
- MOQ 1 per style — the single most valuable term for capsule-brand
economics: zero inventory risk at validation stage
- Transparent unit economics: FOB unit price = verified unit cost +
fixed % supply-chain-management fee (reference: 25%) — never opaque
all-in quotes
- Inspection-gated payment: deposit (50%) up front, balance (50%) only
after final QC inspection before shipment — never 100% prepay
7.3 Partner-evaluation checklist
- Sample tiers match our garment complexity spread (a knit tee ≠ a tailored
jacket — price both)
- Verified-cost access: can we see the mill invoice, or only the bundle?
- QC protocol: who inspects, against what standard, with photos/reports?
- Fabric-lab access for signature/heritage materials (motif design →
handcrafted development → application), with honest lead-time and
advance-booking rules
- Capacity range: does MOQ-1 scale gracefully to capsule-100?
- IP protection on bespoke motifs and patterns
7.4 Skill integration point
When a client brand reaches Phase 1 (prototype), run this checklist on
2–3 candidate partners; the sampling-tier pricing becomes line items in
the brand's financial model (§2.4 margin math assumes FOB + SCM fee, not
FOB alone).
8. Handoff Wiring
Fashion outputs live in the brand's own docs (entity page = law, soul doc =
feeling, design handoff = artifacts). This skill is the method; keep
brand-specific numbers (policies, SKUs, GSMs) in the brand's documents and
link back to this skill only for the how.
Reference implementation artifacts (an instance, not the skill):
technical-flat generator pattern + application spec with rejection codes +
rigorous claims table — see any apparel brand's design handoff in the
vault Handoffs/ directory.
1---2name: fashion-design3description: USE WHEN designing fashion apparel collections as systems: converting a fashion brand proposal into wardrobe systems and fit-block architecture, planning capsules with SKU math, setting pricing/margin guardrails, writing garment spec language (GSM, dry-time, pill cycles), building technical flats for garments, structuring designer application/portfolio specs for fashion hiring, or fashion-specific market/competitor evidence work (in-market validation, competitor store/product analysis). Triggers: "design a capsule", "wardrobe system", "fit blocks", "SKU math", "tech pack language", "garment spec", "technical flat", "fashion application spec", "fabric performance requirements", "tropical/market specific apparel", "hero product strategy". NOT for logo/monogram design (use brand-design), general brand-system orchestration (use brand-system-builder), or non-fashion product design.4---56# Fashion Design — Systems, Capsules & Specs78The craft layer for apparel brand-building: wardrobe architecture, capsule9economics, garment spec language, and fashion-specific evidence discipline.10Brand-agnostic — bring your own brandDNA.1112## 1. Wardrobe System Architecture1314Structure the offer as **2–5 systems named for wearer states, not garment15categories or demographics** (reference pattern: FORM/MOVE/FLOW/TRANSIT/REST —16studio to recovery, one daily arc). Rules:1718- Every SKU belongs to exactly one system; system tag appears on site,19 labels, and internal docs — this is the primary navigation axis20- **Fit logic is a separate axis from garment type.** Define 2–3 fit logics21 (reference pattern: R = one shared pattern / T = same design, blocks A+B22 where anatomy changes the pattern / C = body-specific essentials). The23 naming is an ethical position, not a sizing chart — whichever24 demographics-split convention exists in the category (men/women racks,25 "for her/him"), the brand's own tests must forbid it26- Hero product logic: choose the *least crowded, most describable* garment27 as entry wedge ("one trouser anyone can wear and describe to a friend"),28 versioned like software rather than seasonally replaced2930## 2. Capsule & SKU Math3132- Capsule = designs × colors × size-count; shared patterns multiply cheapest33 (one pattern serves all bodies; per-block patterns multiply by block count)34- Publish the arithmetic in the entity doc: `(shared designs × colors ×35 sizes) + (block designs × blocks × colors × sizes) + … = total SKUs`36- Set a hard design cap (reference: 10 designs first capsule). New styles37 must *retire or version* an old one — SKU inflation is silent and fatal38- Margin guardrails: target GM band at RRP, wholesale floor, and a39 no-launch-below floor; no-discount policy except end-of-line retirement40- Pricing position: anchor explicitly between named competitors, with the41 reason each anchor exists4243## 3. Garment Spec Language (tech-pack register)4445Every garment spec states **measurable claims, not adjectives**:4647| Weak | Spec-grade |48|---|---|49| "ultra-light" | 92 gsm |50| "quick dry" | overnight dry at 80% RH |51| "durable" | 40 wash cycles, no visible pill |52| "comfortable" | air-conditioning transition tolerant (28°C Δ) |5354- Fabric property checklist for tropical/humid markets: wick, dry-time,55 crease, stretch-recover, AC-transition, wash resilience, pill/cling,56 hand-feel (matte, not sport-flash)57- **Label honestly:** proposal-authored numbers are *targets*. Only58 wear-test data upgrades them to claims. Keep a target/validated marker59 next to every metric60- Construction language: invisible-reinforced seams, stress-point bars,61 hardware tonal not branded; the *signature detail* is construction62 (curved waistband, spine seam), never print6364## 4. Technical Flats (drawing garments that match captions)6566Draw flats as pattern-language, not fashion illustration:6768- Off-white ground, ink outlines (1.5–2 units at 400×500)69- Silhouette-specific geometry: trousers (waistband + taper/flare path),70 tees (drop shoulder + rib collar arc), outerwear (collar band + zip line71 + pocket rectangles), knitwear (cuff ribs, drawcord), compression72 (contour curves + gusset lines)73- Always include: grain arrow, 2–3 notches, interior type + drawcord where74 real, seam-path of the brand's construction signature75- Spec line under each flat in mono: fabric + GSM + key property76 (`165 GSM JERSEY · BONDED SEAMS · A/B`) — data only, never slogans77- One Python function per garment type; caption-matching is enforced by78 construction, not by hope7980## 5. Fashion Evidence Discipline8182- **In-market validation beats desk research:** capture real competitor83 store reports (layout, rack split, materials, expansion count) as raw84 notes; map each brand proposal claim → confirmed / contradicted /85 not-evidenced, with differentiation per row86- Track competitor motion via daily intelligence (brand-narrative revamps,87 textile-science waves) and route findings into the entity page's88 competitive-watch section with dates89- Fabric partner sourcing is a Gate-0-class activity: *source fabric before90 design*, not after91- Fit blocks require fit models per block and grading obsession — the92 pattern exists only when a real body in real humidity proves it9394## 6. Fashion Hiring & Application Spec9596When hiring designers for the brand, the application is a design test:9798- Exact-look submission count (forces editing skill — mirror the capsule cap)99- The task is to design *inside the existing system* (11th piece) with100 palette locked, fit logics fixed, SKU math required101- A page arguing *against* the candidate's own design102- Rejection codes (numbered) for auto-disqualification: over-count,103 demographic styling codes, off-brief, oversize SKU, logo-print reflexes,104 no self-critique105- Scored rubric with pass gate + per-section floors106107## 7. Production Partner Integration (on-demand supply chain model)108109Modern indie fashion brands rarely own factories — they integrate an110**on-demand sourcing/SCM partner**. The reference model (verified from111Creabra's published terms, 2026-08) defines the pattern to evaluate any112partner against:113114### 7.1 The capability chain to require115`design brief → [optional fabric lab / heritage textiles] → material116sourcing → sampling → production → QC/final inspection → shipping`117118### 7.2 The commercial structure to expect119- **Tiered sample development** with published fixed prices by complexity120 tiers (reference: standard / structured / complex; specialty fabrics &121 haute couture quoted separately)122- **MOQ 1 per style** — the single most valuable term for capsule-brand123 economics: zero inventory risk at validation stage124- **Transparent unit economics**: FOB unit price = verified unit cost +125 fixed % supply-chain-management fee (reference: 25%) — never opaque126 all-in quotes127- **Inspection-gated payment**: deposit (50%) up front, balance (50%) only128 after final QC inspection *before* shipment — never 100% prepay129130### 7.3 Partner-evaluation checklist131- Sample tiers match our garment complexity spread (a knit tee ≠ a tailored132 jacket — price both)133- Verified-cost access: can we see the mill invoice, or only the bundle?134- QC protocol: who inspects, against what standard, with photos/reports?135- Fabric-lab access for signature/heritage materials (motif design →136 handcrafted development → application), with honest lead-time and137 advance-booking rules138- Capacity range: does MOQ-1 scale gracefully to capsule-100?139- IP protection on bespoke motifs and patterns140141### 7.4 Skill integration point142When a client brand reaches Phase 1 (prototype), run this checklist on1432–3 candidate partners; the sampling-tier pricing becomes line items in144the brand's financial model (§2.4 margin math assumes FOB + SCM fee, not145FOB alone).146147## 8. Handoff Wiring148149Fashion outputs live in the brand's own docs (entity page = law, soul doc =150feeling, design handoff = artifacts). This skill is the method; keep151brand-specific numbers (policies, SKUs, GSMs) in the brand's documents and152link back to this skill only for the *how*.153154**Reference implementation artifacts (an instance, not the skill):**155technical-flat generator pattern + application spec with rejection codes +156 rigorous claims table — see any apparel brand's design handoff in the157vault `Handoffs/` directory.