S2 · Silhouette and fit

Part II teaches the styling knowledge the prototype reasons over; this chapter explains how proportion-related styling enters the agent as user-declared preference, never as classification.

The concept

"Silhouette" in styling literature usually means how a garment's cut and drape interact with the wearer's proportions. The industry convention labels human bodies with shape categories — hourglass, pear, apple, inverted triangle, rectangle — and pairs each with a small set of proportion-related styling rules.

Wearly uses these labels, but with three constraints that make them honest:

1. The labels are user-declared. The agent never analyzes images or measurements to infer a body shape. If a user wants to use the body_shape field, they fill it in themselves.
2. Every field is optional with a safe default. A blank body_shape skips every proportion-related rule. The agent's output never requires a body-shape label.
3. The vocabulary is body-positive by contract. No corrective tokens — flaw, fix, hide, minimize, slimming, problem area — ever appear in a reasoning line or in this documentation. See §S8 · Body-positive framing as a discipline.

The styling action of a shape label is small on purpose. An hourglass preference, when present, surfaces lines like "this silhouette honors the wearer's preferred waist definition" — not prescriptions, not exclusions. The default behavior of the agent if the field is blank is identical to its behavior when the field is present and the wearer declares no proportion concerns.

Where this knowledge comes from

The shape-category vocabulary (hourglass / pear / apple / inverted triangle / rectangle) is industry heuristic, not scientific taxonomy. It descends from mid-20^th-century fashion-magazine and trade-pattern literature and has no agreed academic standing. Wearly uses it because users recognize it and because it's a compact way to encode preferences a wearer already has — not because it's a valid classification of human bodies.

The framing that makes the industry vocabulary usable without harm:

• Hokka (2024) documents body diversity as a known challenge for inclusive design and treats a small set of shape categories as a recognized limitation rather than a universal truth. ⟶ References [7]. The Wearly framing — "if you've shared a body-shape preference" rather than "based on your body type" — is a direct expression of that limitation.

What Wearly deliberately does not do

• It does not infer shape from photos. There is no image classifier, no body-detection model, no measurement extraction.
• It does not enforce body-shape rules. A blank field skips every related rule; a populated field surfaces preference-aware notes but never excludes a garment.
• It does not use corrective vocabulary. The body-positive language contract is a runtime check, not just a documentation aspiration.
• It does not claim the shape vocabulary is a taxonomy. The vocabulary is a proxy for proportion-related preferences; the user is the final reviewer.

How Wearly applies this

Where	What happens
Rule pack	fit-silhouette-rules.md — R1 (body-positive contract), R2 (optional fields), R8 (measurement framing)
Runtime code	fit_tool.fit_alignment_notes(), fit_tool.check_reasoning_for_forbidden_language()
Agent step	Step 2 (load profile) and Step 5 (outfit construction)
Where it surfaces	Fit-alignment notes inside the reasoning trail; the regenerate banner when the profile changes mid-session
Part I chapter	Chapter 06 · Fit-profile logic
MicroSim	(none yet — candidate for a future sim that shows which rules fire when each fit-profile field is set vs. blank)

---

Key terms

• Fit profile — optional, user-declared preferences. Every field has a safe default. See Glossary.
• Body shape (preference) — a user-declared proportion preference, never inferred from images. See Glossary.
• Body-positive language contract — the runtime + documentation rule that forbids corrective vocabulary. See §S8.

Self-check

1. (Remember) List three forbidden tokens from the body-positive language contract.
2. (Understand) Why does Wearly call body_shape a preference rather than a classification?
3. (Apply) A user leaves body_shape blank. Which proportion-related rules fire, and which don't?