user-corpus-explorer

To build a reference index from PDFs and notes, place your files in the external/user-corpus directory. The skill recursively scans the folder, extracts text from PDFs and DOCX files, and generates an AGENTS.md reference library for your math modeling pipeline.

The best way to extract text from PDFs into a local knowledge base is using file-type aware extraction with parsing fallbacks. This skill reads full text from markdown and DOCX files while applying truncation to large PDFs to control processing costs.

Yes, document indexing creates a searchable local reference source of recommendations and limitations. By generating a JSON inventory alongside AGENTS.md, downstream agents can reliably query your user corpus to find relevant papers and examples.

No, the corpus indexing process avoids uploading full documents to external services. It safely builds a local reference index and overwrites the external/user-corpus/AGENTS.md file during each run without blocking the main pipeline on failure.

A limitation of recursive file scanning is handling very large files, which are truncated to control costs. The process applies file-type specific exclusions and parsing fallbacks, but extremely large PDFs may not be fully extracted into the reference index.