kesslerio/mineru-pdf-parser-openclaw-skill
Overview
This skill parses PDF files locally using MinerU on CPU and produces structured Markdown and JSON outputs by default. It can optionally extract tables and images and organizes results into per-document output folders. The skill is focused on reliable local processing without remote services. Outputs are suitable for downstream processing, search indexing, or human review.
How this skill works
The skill runs MinerU CLI on a single PDF and writes results into an output root directory, with MinerU creating a subfolder per document (e.g., ./mineru-output/<basename>/). By default it emits Markdown plus a JSON manifest; optional flags enable table and image extraction and alternate formats. It operates on CPU and uses local resources and parameters supplied to the wrapper script for format and extraction choices.
When to use it
- You need offline PDF parsing on a CPU-only machine.
- You want Markdown and JSON outputs organized per document for downstream tooling.
- You require optional extraction of tables and images from PDFs.
- You prefer a simple CLI wrapper to standardize MinerU invocations.
- You are processing individual PDFs rather than large batches.
Best practices
- Run the parser from the skill directory or set the working directory so output paths are consistent.
- Specify --tables and --images only when you need those assets to avoid extra processing time and storage.
- Keep the output root organized and back up or archive per-document folders regularly.
- Tune MinerU flags for device, threads, or format if you need different performance or precision.
- Validate the JSON manifest after processing to confirm expected assets and metadata were produced.
Example use cases
- Convert a research paper PDF to Markdown and a JSON summary for a content ingestion pipeline.
- Extract tables and images from a financial report into discrete files for analysis.
- Normalize client-delivered PDFs into a per-document folder structure for archival.
- Run quick local parsing for privacy-sensitive documents without using cloud services.
- Prepare legal documents into Markdown for easier review and citation management.
FAQ
Not by default; the wrapper focuses on single-PDF parsing. Implement batch folder parsing only if explicitly required.
Where are outputs stored?
Outputs go under a configured output root (default ./mineru-output/) and MinerU creates a subfolder per document using the PDF basename.
