prep-photo

Prep Photo

Pre-process a hardware photo to improve clarity and alignment for visual component identification. The skill invokes a Python script to auto-rotate based on EXIF, de-skew if needed, optionally enhance contrast and white-balance, and crop to the board boundary. Output is a cleaned-up image saved with a -prepped suffix.

Data storage root:

${CLAUDE_USER_DATA:-${XDG_DATA_HOME:-$HOME/.local/share}/claude-plugins}/hardware-id-annotation/

When to use

• Photo is captured at an angle or portrait orientation and needs straightening
• Image has poor lighting, glare on components, or low contrast silkscreen
• You want to auto-crop to just the board (remove hands, background clutter)
• Original photo is not ideal for identify-components

Inputs to gather

• Input image — JPG or PNG of hardware
• Output directory — optional; defaults to same directory as input or output/
• Processing flags — any combo of:
  • --auto-rotate (default: true) — use EXIF orientation
  • --deskew (default: false) — rotate to align board edges (requires opencv)
  • --enhance-contrast (default: false) — boost visibility of faint markings
  • --white-balance (default: false) — correct color cast
  • --crop-board (default: false) — auto-detect and crop to board boundary (requires opencv)

Procedure

1. Invoke the prep script:

   python3 ${CLAUDE_PLUGIN_ROOT}/scripts/prep.py \
     --input <image> \
     --output <output.png> \
     --auto-rotate \
     --deskew \
     --enhance-contrast \
     --white-balance
   

2. Inspect the output. Display the prepped image to the user and ask: "Does this look better? Any areas still unclear?" If needed, suggest running again with different flags (e.g., less contrast boost, crop enabled).

3. Save prepped image with -prepped suffix by default:

   <input-dir>/<name>-prepped.png
   

   or to plugin storage if specified:

   ${CLAUDE_USER_DATA:-${XDG_DATA_HOME:-$HOME/.local/share}/claude-plugins}/hardware-id-annotation/output/<board-slug>-prepped.png
   

4. Preserve original. Never overwrite the original unless explicitly confirmed by the user.

5. Report transformations. Console output lists which operations were applied and their parameters (e.g., "Rotated 90° CCW per EXIF, deskew 3.2°, contrast +20%").

Output / side effects

• Prepped PNG file saved to disk with -prepped suffix
• Metadata: original filename, transformations applied, file size comparison
• Console: before/after dimensions, auto-rotation angle, deskew angle (if applied), processing time
• User can now pass the prepped image to identify-components for cleaner component detection

Safety / constraints

• EXIF rotation is always attempted but safe to skip if EXIF is absent.
• De-skew and crop require opencv-python, which is optional; script should gracefully skip or warn if not installed. Use install-deps to add it.
• Contrast and white-balance are destructive; preview carefully before proceeding. Offer undo or re-run with milder settings.
• Lossy compression — output is PNG (lossless) by default, but user can override to JPG at quality 95 if file size is a concern.
• Large images (4K) may take 5–10 seconds; inform user upfront.