ocr

name	ocr
description	Run local OCR on an image file using simpleocr. Use when the user wants to extract text from an image, read a screenshot, analyze a document, or get structured data from a photo.
argument-hint	<image-path> [--format plain|text|json|table-json]
allowed-tools	Bash(simpleocr *)

Run OCR on the specified image using simpleocr.

Important: simpleocr only works on macOS (it uses Apple's Vision framework). If the user is not on macOS, inform them that this skill is not available on their platform.

Usage

/ocr <image-path> [options]

Steps

1. Run: simpleocr $ARGUMENTS
   • If no --format is specified, use --format plain (best for LLM consumption).
   • If the command fails with "command not found", tell the user to install simpleocr via brew install tobilg/simpleocr/simpleocr.
2. Show the OCR output.
3. Briefly summarize what was recognized (number of text blocks, document type if obvious, any tables detected).

Available formats

• plain — one line per text element, sorted top-to-bottom then left-to-right (default, best for LLMs)
• text — same as plain but with normalized [y=,x=] coordinates prepended
• json — full JSON with observations, bounding boxes, confidence scores, and structured regions
• table-json — JSON with only table-like structured regions and row/cell data

Additional options

• --lang <codes> — comma-separated language codes (default: de-DE,en-US)
• --mode fast — faster but less accurate recognition
• --pii — redact personally identifiable information
• --min-confidence <val> — minimum confidence threshold 0.0-1.0 (default: 0.3)
• --error-format json — machine-readable error output

Examples

/ocr screenshot.png
/ocr invoice.pdf --format json
/ocr photo.heic --format table-json --lang en-US
/ocr receipt.jpg --pii --format plain