Eyeglasses Selection

---
name: eyeglasses-selection
description: Compare prescription eyeglass frame shapes, materials, sizing and everyday aesthetics using consented photos or descriptions; assess existing-frame fit, build ranked shortlists, make controlled virtual try-on previews, and prepare optician-ready fitting questions. Use for selecting glasses, comparing navigator/panto/rectangular frames, virtual eyewear try-ons, or explaining frame choices alongside an existing spectacle prescription.
metadata:
  version: "1.0.0"
---

# Eyeglasses Selection

## Goal and limits

Support informed selection of **eyeglass frames**, separating visual suitability, physical fit, and optical prescription requirements. Frames can be visually flattering and still fit badly. Do not present an AI image as evidence of precise lens thickness, pupillary centring, visual acuity or wearing comfort. A qualified optician/optometrist performs actual dispensing and refraction.

The skill applies to general style exploration, comparing specific frame models, current-frame assessments, and user-requested virtual try-ons. It does not issue, revise or infer a spectacle prescription or diagnose worsening vision.

## Intake without repetitive questions

1. Identify the immediate decision: **explore frame families**, compare **specific purchasable models**, evaluate **current frames**, or prepare for an **in-person try-on**. A frame-style request can proceed without collecting medical information.
2. If the person has a comfortable, secure existing pair, treat it as a **physical-fit baseline**, not as automatic proof that its style is best.
3. Gather only available preference/fit facts: intended use (primary vs secondary), visual impression, current frame shape and comfort, optional budget, hair/eyebrows/bridge fit, willingness to change shape, and confirmed prescription status if relevant.
4. Photos must be provided or authorised for that processing. Minimise background, metadata and unnecessary identifiers. Preserve privacy when working with somebody else's portrait. Do not assume metadata stripping anonymises an identifiable face.
5. If the current prescription is missing or not final, compare frame aesthetics provisionally and explain that lenses must wait for professional verification. Ask a focused question only if its absence prevents the requested decision.

## Separate three independent frame dimensions

- **Shape:** rectangular, soft-square, D-frame, panto, round, oval, browline, squared navigator, teardrop aviator, subtle geometric, etc.
- **Construction:** full-rim, semi-rimless or rimless.
- **Material/colour:** acetate, titanium, steel, polymers and combinations; solid, translucent, tortoiseshell, neutral colours.

Do not call "titanium," "rimless," and "aviator" mutually exclusive types. Explain style families first, then how rim and material refine them.

Read [frame taxonomy](references/frame-taxonomy.md) when classifying frame designs and [optical fitting](references/optical-and-fit.md) when discussing prescription, measurements, bridge, lenses or discomfort.

## Comparison and ranking workflow

1. **Baseline:** State what is already supported by real-world experience (e.g., existing pair fits all day). Note any observed strengths/weaknesses rather than relying on face-shape clichés.
2. **Style exploration:** For open-ended requests, compare about 8–10 visibly distinct everyday frames, including a justified example of the current-frame family when available. For specific product shopping, use exact model images/dimensions and current listings from appropriate available tools; do not invent products, prices or stock.
3. **Facial and outfit context:** Consider apparent face width/height, eyebrow-frame interaction, eye position, cheek clearance, nose/bridge fit, hairstyle volume, colour harmony, daily scenarios and taste. Photo-based geometric claims remain approximate.
4. **Optical practicality:** Prefer reasonable lens size and centring for substantial minus prescriptions, but do not declare a model unsuitable simply because of lens shape. Larger minus lenses can have thicker edges; actual thickness depends on full Rx, dimensions, centring, lens material and lens design.
5. **Fair shortlist:** Rank about three finalists against consistent criteria—visual balance, everyday versatility, fit evidence, lens practicality, and user preference. Name pros, drawbacks, and why a different model could outrank the current pair. Avoid fabricated numerical attractiveness scores.
6. **Decision:** Aesthetic choice may be made now; final frame purchase and prescription lens ordering require actual fitting and a **verified final prescription**.

## Optical boundary

- Read `SPH`, `CYL`, `AXIS`, `OD/OS` and PD exactly as supplied. Do not assume missing +/− signs, approximate machine readings, or eye order. Automated refractor printouts are **not necessarily the final spectacle prescription**.
- Monocular PD, optical centre height, bridge fit, vertex distance, pantoscopic tilt, temple length and frame adjustment may require professional measurements. PD alone does not yield a guaranteed frame size.
- A higher lens refractive index is not inherently a better purchase; compare finished thickness, weight and visual comfort with the optician.
- When the user reports worsening astigmatism or other changing vision, recommend a comprehensive eye assessment before ordering new prescription lenses. Sudden vision change, pain, flashes, or acute symptoms warrant prompt professional care; do not diagnose a condition from values or portraits.

Read [optical and prescription precautions](references/optical-and-fit.md) if any prescription or lens advice is requested.

## Visual try-on protocol

If the user explicitly asks for generated images and an appropriate tool exists, follow [preview production](references/preview-and-delivery.md): use one anchored photo and change only frame design; evaluate all candidates in comparable front views before adding matched oblique and lateral views for finalists. Don't present invented frame dimensions as exact model data.

For a general-style comparison, choose plausible medium-sized reference frames and label them **illustrative**, not commercially exact. For real products, preserve the actual supplied frame image/details instead of substituting generic renderings. Do not make a user upload more identifiable portraits than necessary.

## Output contract

Match the user's requested scope. A full consultation should provide:

1. **Type map** (shape vs construction vs material) when requested.
2. **Observed context and limitations** (incl. what is known about the current frame).
3. **Ranked frame candidates** and clear style/fit/trade-off rationales.
4. **Optional user-requested try-ons**, explicitly conceptual and visually quality-checked.
5. **In-person try-on and dispensing checklist**: actual bridge/cheek/temple comfort; frame sizing; optician-confirmed prescription and lens planning.
6. **Recommendation status:** provisional until actual frame try-on and optical fit checks.

## Quality gate

Check subject consent, reused constraints, type taxonomy, accuracy of Rx notation, product fact provenance, frame style distinctiveness, no induced face reshaping, consistent camera reference, limits of image-based lens fit, and no false promise of medically verified prescription performance. Ask one decision-changing question if truly required; otherwise proceed transparently.