PDF to Markdown with OCR

Document OCR and parsing — converts PDF/images to Markdown on remote L4 GPU via Modal. Trigger when user says: OCR, PDF to markdown, parse PDF, extract text...

This skill appears to implement exactly what it claims: it uploads local PDFs to Modal volumes, runs an OCR pipeline (mineru) on a remote GPU image, and downloads Markdown outputs. Before installing: (1) ensure you trust the mineru package and the container image (it will pip-install mineru inside the remote image); (2) understand that it will create/use Modal volumes named speech2srt-data and speech2srt-models in your Modal account — these are shared/account-level resources and may already contain or be used for other data; (3) the pipeline symlinks the runtime ~/.cache into the models volume (it will remove an existing cache directory in the runtime), so check for collisions with any existing cached content you care about; (4) the skill requires a Modal account and may consume paid GPU credits, so verify billing/credits before running. If you need stronger isolation, change the volume names and review the image/pip packages used.

Type: OpenClaw Skill Name: ocr2markdown Version: 1.0.1 The skill bundle implements a legitimate OCR pipeline using the 'mineru' library on the Modal serverless platform. The code in 'src/ocr2markdown.py' and 'src/images.py' follows standard Modal patterns for GPU-accelerated workloads, including symlinking cache directories to persistent volumes for model storage. No evidence of data exfiltration, malicious execution, or harmful prompt injection was found; the instructions in 'SKILL.md' are consistent with the stated purpose of processing PDF files.