V3 deterministic searchable-PDF and OCR-text capability pack. By-value connector composition (service_kind=cli, no separate ingredient): one local OCRmyPDF CLI ingredient exposes bounded operations for readiness, searchable-PDF generation, and OCR sidecar text extraction. Bundles a direct searchable-PDF recipe plus a scanned-document notes workflow that chains OCRmyPDF with PDF text extraction and preview-first AI summarization. Fixed-function and reproducible: the executor materializes the source file and owns the throwaway output directory, while OCRmyPDF writes fixed captured outputs. Write-tier but approval=never: OCRmyPDF writes only to engine-managed output dirs, exposes no arbitrary plugin/config flags, no caller-supplied output paths, no deskew/clean/rotate mutators, no force-ocr, no Tesseract config overrides, and source bytes never flow through op-step values. Requires ocrmypdf plus Tesseract language data on PATH (Debian/Ubuntu: apt install ocrmypdf tesseract-ocr-eng). Local and no-egress.