Scored across 28 tools
Most tools have sharply distinct purposes, and the descriptions actively resolve the obvious overlaps: generate_basic_pdf is declared the default while add_table/add_chart/embed_image/add_barcode are scoped to standalone artefacts, split_pdf vs extract_pages is cross-referenced, and add_form vs read_form_fields vs fill_form are clearly separated. The only soft spot is the cluster of read-only inspectors (inspect_pdf, inspect_layout, validate_pdf, verify_pdf), whose names alone would be confusable, though their descriptions disambiguate well.
All names are snake_case with a leading verb (generate_basic_pdf, extract_text, merge_pdfs, sign_pdf, validate_pdf, verify_pdf), which is predictable. The main deviation is the create-family split between generate_/embed_ and the add_* prefix (add_table, add_form, add_chart, add_barcode, add_attachment, add_ltv), plus a few oddities like timestamp_pdf and prepare_signature_placeholder.
At 28 tools this is on the heavy side, well beyond the comfortable 3–15 band. The PDF domain is genuinely broad (generation, extraction, page ops, forms, signatures, LTV, validation), so each tool does have a distinct job, but the surface is large enough that agents must scan a long list to find the right entry point.
Coverage is unusually thorough: creation from blocks, standalone artefact builders, extraction, merge/split/subset, annotation, encrypt/decrypt, form read/fill, signing with placeholder, DSS/LTV, document timestamps, structural validation and signature verification form a coherent lifecycle. Remaining gaps are true redaction (explicitly called out as out of scope for annotate_pdf) and page-level editing like rotation/reordering in place.