S1 · Color and skin tone

Part II teaches the styling knowledge the prototype reasons over; this chapter explains the color framework Wearly applies at Step 7.

The concept

Wearly's color logic rests on the warm / cool axis — the design-school convention that classifies skin undertones and garment colors along the same warm-to-cool dimension and then asks how well a color sits with the wearer. The convention is centuries old; Wearly's implementation is the rule layer that operationalizes the convention against an actual wardrobe.

Three named palettes ship today:

• warm olive — yellow / golden undertone with green influence.
• cool fair — pink / blue undertone with low contrast.
• deep warm — rich brown / golden undertone with high contrast.

Each palette declares three tiers — best, good, avoid — and the agent scores an outfit against the wearer's palette:

score = 60                                  # neutral base
for each item:
    if item.color in palette.best:    score += 8
    elif item.color in palette.good:  score += 4
    elif item.color in palette.avoid: score −= 10
score = clamp(score, 0, 100)

Three lookups, one clamp — every term visible in the reasoning trail. The +8 / +4 / −10 asymmetry is a deliberate UX choice: a clash should cost more than a match adds, so the score is conservative.

Where this knowledge comes from

Color theory is not a science of "correct" palettes, but it does have a recognized vocabulary and a measured foundation:

• The warm / cool framework descends from Itten's Elements of Color — the formal vocabulary of color contrasts (hue, light/dark, cold/warm, complementary, simultaneous, saturation, extension) that design-school color teaching builds on. ⟶ References [4]
• The measured-color foundation under named palettes goes back to Munsell (1905), who formalized hue × value × chroma as a system a century before today's named palettes. ⟶ References [5]
• The modern perceptual standard is the CIELAB color space (ISO 11664-4:2008). Wearly does not compute ΔE distances at runtime; the citation acknowledges the framework Wearly's named palette sits above. ⟶ References [6]

What Wearly deliberately does not claim:

• That its palettes are empirically optimal for any wearer. They're design-informed conventions, not findings from a randomized study.
• That seasonal-color analysis ("Color Me Beautiful" and similar industry frameworks) is science. The framework is widely referenced but not peer-reviewed; Wearly cites it as industry practice, never as evidence.

Why named colors instead of CIELAB

Three reasons, in order of importance:

1. The rule has to be auditable. A reasoning line that reads "forest green pairs well with deep-warm undertones" is a sentence a non-expert user can argue with. "ΔE 12.4 against L* 43.2" is not.
2. The data is small. Three palettes × roughly 30 named colors each fits in color_rules.json and is editable by a non-engineer.
3. The honesty contract demands it. "Tends to work well" is a calibrated claim; a numeric distance suggests precision Wearly does not have.

The trade-off: Wearly cannot reason about colors it doesn't have a name for, and it cannot compute contrast between two named colors in the same outfit. Both are deliberate scope choices — see Chapter 04 for the "why rules, not ML" discussion.

How Wearly applies this

Where	What happens
Rule pack	color-coordination-rules.md — R1 (palettes), R2 (scoring), R5 (unknown skin tone)
Runtime code	color_tool.score_outfit_colors()
Agent step	Step 7 of the seven-step workflow
Where it surfaces	The color-score panel + the per-item +8 / +4 / −10 note in the reasoning trail
Part I chapter	Chapter 04 · Styling knowledge base
MicroSim	Color harmony — try the formula on every combination

---

Key terms

• Skin-tone palette — one of three named palettes (warm olive, cool fair, deep warm). See Glossary.
• Color tier — best (+8), good (+4), avoid (−10). See Glossary.
• Warm / cool axis — the design-school dimension the three palettes pivot on.

Self-check

1. (Remember) What are the three palette tiers and their numeric contributions?
2. (Understand) Why does Wearly use named colors instead of CIELAB distances at runtime?
3. (Apply) For a cool-fair wearer, predict whether navy + cream + cobalt lands above or below 80, and justify your answer in one sentence.

Self-check follows Bloom's progression — same pattern as Part I chapters and MicroSim pages.