passport-photo-check

passport-photo-check

A Claude Code skill that verifies every individual photo in a print-lab layout JPEG meets official ID/passport requirements. Given a 4×6 (or other) tiled layout, it runs both a visual compliance check (face proportion, expression, background, gaze) and a precise pixel-level dimension measurement. Pairs with /photo-print-layout, which produces the layout file this skill checks.

What It Does

• Loads the layout JPEG and visually inspects each photo against the target document's official checklist
• Detects content boundaries by scanning for non-white pixels, then reports each photo's exact width and height in millimeters
• Covers Chinese passport, US passport, UK biometric, and other common ID photo formats
• Flags the most common failure: face too small (does not occupy the required fraction of photo height)

Workflow

flowchart TD
    A(["/passport-photo-check\nlayout.jpg + doc type"]) --> B[Load layout JPEG\nvia Read tool]
    B --> C[Visual inspection\nper-photo per-criterion]
    C --> D{Each photo\npasses checklist?}
    D -->|Issues found| E[Flag: face proportion\ngaze · background · expression\nglasses · lighting]
    D -->|All pass| F[Visual: PASS]
    E --> F
    B --> G[Get DPI via sips\ndefault 300]
    G --> H[Run PIL pixel scan\nfind content boundaries]
    H --> I[Detect all-white column gaps\n→ photo widths in mm]
    I --> J[Detect all-white row gaps\n→ photo heights in mm]
    J --> K{Dimensions match\nofficial spec?}
    K -->|Yes| L[Dimension: PASS]
    K -->|No| M[Dimension: FAIL\nreport delta]
    F --> N[Final summary\npass · fail · action items]
    L --> N
    M --> N
    N --> O{Any failures?}
    O -->|Yes| P["Action: re-crop source\nretake photo\nre-run /photo-print-layout"]
    O -->|No| Q(["Done\nReady to print"])

Loading

Install

git clone https://github.com/biomystery/claude-skills.git
mkdir -p ~/.claude/skills
ln -s "$(pwd)/claude-skills/passport-photo-check" ~/.claude/skills/passport-photo-check

Restart Claude Code — /passport-photo-check will be available.

Usage

# Interactive — Claude asks for layout path and document type
/passport-photo-check

# Chinese passport layout
/passport-photo-check ~/passport_4x6_walgreens.jpg chinese-passport

# US passport layout
/passport-photo-check ~/us_passport_4x6.jpg us-passport

Output

Sample output (illustrative values):

Full image: 1200x1800px = 101.6mm x 152.4mm at 300 DPI

Dimension check:
  Col 1: 390px = 33.0mm  Col 2: 390px = 33.0mm
  Row 1: 567px = 48.0mm  Row 2: 567px = 48.0mm  Row 3: 568px = 48.1mm
  → All photos: 33mm × 48mm ✓  (required: 33 × 48mm)

Visual check:
  Row 1: face ~55% of height → FAIL (need ≥58%)
  Row 2: face ~60% → borderline
  Row 3: face ~65% → PASS
  Background · expression · gaze · lighting: PASS across all rows

Overall: FAIL — face too small in rows 1–2.
Action: re-crop source photo tighter, re-run /photo-print-layout.

Supported Document Types

Document type	Photo size (mm)	Face height requirement	Background
Chinese passport / visa	33 × 48	28–33mm (58–69%)	White
US passport	51 × 51	25–35mm (49–69%)	White or off-white
UK biometric passport	35 × 45	29–34mm (64–76%)	Light grey or cream
Indian passport	35 × 45	≥ 70% of frame	White
1-inch 一寸 (China)	25 × 35	same as Chinese passport	White
2-inch 二寸 (China)	35 × 49	same as Chinese passport	White

Requirements

• Python 3
• Pillow: pip3 install Pillow
• Layout JPEG at 300 DPI (standard photo lab output)

Supported Inputs / Edge Cases

• 2×3, 2×2, or other grids: script auto-detects columns and rows from white-gap analysis
• No gap between photos: gap list is empty — reports total content width / column count
• Non-300 DPI files: sips reads actual DPI from metadata; override DPI variable accordingly
• HEIC source: convert first — sips -s format jpeg photo.heic --out photo.jpg
• Portrait or landscape canvas: orientation-agnostic; reads pixel dimensions directly

Skill Structure

passport-photo-check/
├── SKILL.md    (skill definition — what Claude executes)
└── README.md   (this file)