Screenshot OCR Text Extractor

Problem

Text in screenshots, photos, or system UIs is often hard to copy, especially in tables, multi-column layouts, or low-contrast images. Manual transcription can drop words, merge columns, or break hierarchy. Engineers usually need editable plain text they can paste into docs, notes, or code blocks.

How It Works

This skill turns screenshot OCR into a predictable workflow rather than a one-shot image description:
- Image intake: checks image type and confirms whether it contains tables, columns, code snippets, or dense annotations.
- OCR extraction: reads text directly when the platform supports image input; otherwise suggests local tooling such as tesseract or paddleocr.
- Layout proofing: restores paragraphs, removes noise, and fixes obvious misreads, line breaks, or ordering issues.
- Text output: returns text in the conversation by default and does not silently write local files; file saving requires the appropriate file access permission.

Boundaries

It processes locally by default and does not fetch external resources online. It avoids confidential or private screenshots and does not alter the source text. It is useful for first drafts, material organization, and format conversion, but policy decisions and professional proofing still require human review.