Drawing Markup Skill

# Quick extraction of all unit IDs from a PDF
pdftotext -layout "drawing.pdf" - | grep -oE 'B[1-4]\.\d+\.\d+'

Step 2: Extract Type Codes

Type codes are usually positioned near the unit ID in the drawing. Extract them with:

# Raw text extraction preserves spatial proximity
pdftotext -raw "drawing.pdf" - | grep -E '(B[1-4]\.\d+\.\d+|[A-Z]*\.?[1-3]B[12]?B)'

Common type code formats:

• 1B1B — 1 Bedroom, 1 Bathroom
• 2B1B — 2 Bedroom, 1 Bathroom
• 3B1B1P — 3 Bedroom, 1 Bathroom, 1 Powder Room
• 2B2B — 2 Bedroom, 2 Bathroom
• 4B2B — 4 Bedroom, 2 Bathroom

Prefixes: DDA. = accessible, F. = facade variant, TH. = townhouse Suffixes: .M = mirror layout

Step 3: Build Unit-to-Type Mapping

Create a dictionary mapping each unit ID to its type code. Two approaches:

A) From PDF proximity matching — Match unit IDs to the nearest type code below them in the PDF text:

# Type code should be within ~80pt below and ~40pt horizontally of unit ID
if dy > 0 and dy < 80 and dx < 40:
    # This type belongs to this unit

B) From a pre-built register — If you've already compiled an apartment register (e.g., via the xlsx skill), use that mapping directly. This is more reliable.

For typical floors (where one drawing represents multiple levels), implement fallback matching — if a unit ID's level number doesn't match exactly, try the base level of the typical floor range.

Step 4: Generate Markup with the Script

Use the bundled scripts/markup_plans.py script. It takes a JSON config file:

python3 ~/.claude/skills/drawing-markup/scripts/markup_plans.py config.json

Config file format:

{
  "input_dir": "/path/to/pdf/folder",
  "output_path": "/path/to/output/Marked_Up_Plans.pdf",
  "unit_pattern": "B[1-4]\\.\\d+\\.\\d+",
  "type_map": {
    "B1.04.01": "1B1B-01",
    "B1.04.02": "DDA.2B1B-01",
    "B2.04.01": "F.3B1B1P-01"
  },
  "level_names": {
    "LEVEL 04": "Level 04",
    "LEVEL 07": "Levels 07-11 (Typical)"
  },
  "typical_floor_bases": [7, 12, 16, 20],
  "title": "PROJECT NAME"
}

If no config file exists, generate one by:

1. Scanning all PDFs in the input directory for unit IDs
2. Extracting type codes via proximity matching
3. Building the type_map automatically
4. Writing the config for the user to review/edit before final markup

Step 5: Open and Verify

Open the output PDF and verify the markups look correct. Check:

• All units have colored badges (no grey "?" badges)
• Bedroom counts match the type codes
• Legend is visible and not overlapping drawings
• Badges don't obscure the architectural drawing underneath

Annotation Style

The markup uses minimal, non-destructive annotations:

• Colored pill badge above each unit label showing bedroom count (e.g., "1B", "2B", "3B")
• Colored underline beneath the unit ID text
• Legend in the top-right corner of each page
• Level title bar in the top-left showing level name and apartment count
• No large rectangles or opaque overlays that obscure the drawing

Color Scheme

Adapting to Different Projects

The unit ID pattern (B[1-4]\.\d+\.\d+) covers the most common Australian residential convention. For other conventions:

• Update unit_pattern in the config to match the project's naming scheme
• The type code classification logic handles any XB[Y]B[Z]P format automatically
• Add custom prefixes (DDA, TH, F, etc.) as needed to the classifier