format-specific-extraction

Open the file as a ZIP archive, read parts like word/document.xml or ppt/slides/*.xml, and parse them with a streaming quick-xml Reader. Thread a SecurityBudget through the recursive walkers so hostile documents exhaust a budget instead of memory.

Add one FormatEntry to the FORMATS registry in core/mime.rs, implement InternalDocumentExtractor with supported_mime_types and priority, register it in register_default_extractors, and apply SecurityBudget to user content. Feature-gate optional formats and add fixture-based tests.

Yes, pages with no extractable text route to OCR automatically. You can force OCR for the whole document with config.force_ocr or for specific pages with config.force_ocr_pages, and choose between native and pdfium backends.

ZipBombValidator validates the archive against configured security limits before any extraction occurs. Office formats instead use SecurityBudget with a max_files_in_archive member cap, and email nesting depth is bounded by DepthValidator.

No, SecurityBudget, SecurityLimits, ZipBombValidator, DepthValidator, and StringGrowthValidator are pub(crate). In-crate extractors can use them, but out-of-crate plugin authors cannot access these security types.