- Home
- Skills
- Productivity
- Searchable PDF OCR & Text Extraction (macOS)
Works with the AI Agents you already use
Searchable PDF OCR & Text Extraction (macOS)
Add searchable text to scanned PDFs and extract Markdown locally on macOS, with batch processing, preservation presets, and OCR box repair.
Free
Searchable PDF OCR & Text Extraction (macOS)
Example session with this skill installed
On my Mac, use pdf-ocr-searchable to check whether sample-scan.pdf needs OCR. If it is a scanned PDF, create a searchable PDF and Markdown, preserve the original, and report the checks performed. Use the default archive preset and do not use --force.
- Read your context and instructions
- Compiled the searchable pdf ocr
- Generated the document
Illustrative workflow, not an OCR test result
The filename below is an example. No source PDF was processed to produce this listing, so this illustration makes no claim about recognition accuracy or successful output.
Check, then process
After installing the complete local package and dependencies, the agent can run:
S="$HOME/.agents/skills/pdf-ocr-searchable/scripts/ocr_pdf.sh"
"$S" --check "sample-scan.pdf"
"$S" --md "sample-scan.pdf"
For a scan that needs OCR, normal outputs are sample-scan.ocr.pdf and sample-scan.md beside the retained original. If the source already has a usable text layer, the workflow can skip OCR and extract Markdown directly. Existing newer outputs may be reused.
What the final report must verify
Confirm the actual input and output paths, page count, selectable/copyable text, and representative body passages against the original pages. Check complex headings and low-text pages separately. Keep notes of recognition and layout errors. A successful command exit is not evidence that every page is accurate.
Preservation boundary
The default preset avoids re-encoding scanned images, but rewrites the PDF container. Keep the source PDF. Deskew, compression, force, and repair options have different effects and require the corresponding checks in the documentation. Encrypted and already signed files are outside the supported red-box rewrite path.
ocr.pdf
PDF · document
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
Turn scanned PDFs on a Mac into searchable documents with selectable, copyable text, and extract Markdown when needed. OCR runs locally through Apple Vision / Live Text using OCRmyPDF and the AppleOCR plugin. The skill can guide an agent or be used through the included command-line scripts.
Features include PDF triage before OCR, batch and incremental processing, Markdown extraction, language and recognition-mode selection, and archive, deskew, and compact presets. The default archive preset preserves scanned images without re-encoding, although the PDF container is rewritten. Supported AppleOCR debugging boxes can be removed automatically or repaired in existing outputs.
Requirements: macOS, Homebrew, uv, OCRmyPDF with the AppleOCR plugin installed in the same Python environment, Ghostscript, and Poppler. Live Text depends on system support; accurate mode can be used when needed. These tools and runtime dependencies are not bundled. Install the complete skill locally before using the scripts; marketplace instruction loading alone does not supply the runtime or local package paths.
Supported recognition languages include Simplified and Traditional Chinese, English, Japanese, and Korean, subject to the selected mode and system. Combined language strings such as chi_sim+eng are not supported. OCR and reconstructed Markdown can contain errors: spot-check important content against the original pages. Low-text pages and complex layouts need particular care. Red-box repair does not rewrite encrypted or already signed PDFs.
Outputs normally use a new .ocr.pdf name, preserving the original. Force and repair operations can rewrite their target output, so retain the source and follow the included guidance. The package includes the OCR entry script, Markdown extraction and repair scripts, compatibility notes, tests, bilingual README files, installation guides, and the MIT license.
Documentation, installation instructions, and a complete ZIP are available without GitHub at https://www.lumenghe.com/en/skills/pdf-ocr-searchable/ . Chinese documentation: https://www.lumenghe.com/skills/pdf-ocr-searchable/ .
Documentation version: 1.0.1. Verified source commit: bed2dc83e0121a37a306860e90d6071b3e0815f9. Packaged 2026-10-24; this does not assert that a corresponding GitHub Release was verified.
MIT licensed; preserve the included copyright and license notices.
macOS-only workflow with separately installed OCR dependencies. Recognition and Markdown reconstruction can contain errors; spot-check important passages against source pages. Combined language strings such as chi_sim+eng are unsupported. Red-box repair does not rewrite encrypted or already signed PDFs. Image preservation does not guarantee byte-identical PDF files. The demo is illustrative and no PDF was processed for this listing.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
1 use
Times used to date
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 24 days ago
- Free with an Agensi account
Requires
Permissions
Reads user-selected local PDFs and writes OCR/Markdown outputs. Shell scripts use normal HOME/PATH values and an optional OCRMYPDF executable override. OCR and extraction run locally after dependencies are installed; no external OCR service or API key is required. Initial dependency installation needs network access separately. Force or repair options may rewrite their output target; originals should be retained.