PDF Workflow
Inspect first
Record size, page count, version, encryption, permissions, signatures, page boxes/rotation, text-layer quality, fonts, images, attachments, forms, annotations, and whether pages are scanned. Use installed tools such as a PDF parser, renderer, OCR engine, or office converter; do not install dependencies implicitly.
Route by task
- Read/summarize: extract text by page and visually inspect pages containing figures, tables, formulas, or suspicious extraction.
- Tables: compare text extraction with rendered layout; do not treat a visually aligned table as reliable rows until validated.
- Scans/OCR: preserve the original image, record language and OCR settings, and mark uncertain text.
- Merge/split/rotate: preserve page order, boxes, bookmarks, metadata, links, and forms when supported.
- Create: use a layout engine appropriate to the source, embed or license fonts correctly, and add document metadata/accessibility structure where supported.
- Forms: detect AcroForm versus XFA, preserve field names and appearances, and flatten only when requested.
- Redaction: remove underlying text/images and metadata, then verify by extraction and object inspection. Drawing a black rectangle is not secure redaction.
Safety
- Write to a new output path and keep a backup.
- Do not bypass passwords or permissions without authority.
- Warn before an edit that invalidates a digital signature.
- Do not execute JavaScript, launch actions, attachments, or embedded content.
- Remove hidden metadata/attachments only when requested; they may be evidentiary.
- Do not send private PDFs to external OCR or conversion services without explicit authorization.
Validation
Reopen the output, verify page count/order/rotation, render all affected pages, inspect text and form fields, check links/bookmarks/attachments, and confirm no corruption. For secure redaction, search extracted text, inspect content streams where possible, and test copy/paste. Report exact pages changed and any unsupported feature.
1---2name: pdf-workflow3description: Apply a provenance-first PDF workflow for inspecting, extracting, creating, combining, splitting, rotating, redacting, OCR, or form filling. Use when the user asks for this workflow or needs an additional PDF safety and validation checklist alongside the installed PDF skill. Turkish triggers: PDF oku veya oluştur, sayfaları görsel kontrol et, metin/tablo çıkar, PDF düzenle.4license: MIT5---67# PDF Workflow89## Inspect first1011Record size, page count, version, encryption, permissions, signatures, page boxes/rotation, text-layer quality, fonts, images, attachments, forms, annotations, and whether pages are scanned. Use installed tools such as a PDF parser, renderer, OCR engine, or office converter; do not install dependencies implicitly.1213## Route by task1415- **Read/summarize:** extract text by page and visually inspect pages containing figures, tables, formulas, or suspicious extraction.16- **Tables:** compare text extraction with rendered layout; do not treat a visually aligned table as reliable rows until validated.17- **Scans/OCR:** preserve the original image, record language and OCR settings, and mark uncertain text.18- **Merge/split/rotate:** preserve page order, boxes, bookmarks, metadata, links, and forms when supported.19- **Create:** use a layout engine appropriate to the source, embed or license fonts correctly, and add document metadata/accessibility structure where supported.20- **Forms:** detect AcroForm versus XFA, preserve field names and appearances, and flatten only when requested.21- **Redaction:** remove underlying text/images and metadata, then verify by extraction and object inspection. Drawing a black rectangle is not secure redaction.2223## Safety2425- Write to a new output path and keep a backup.26- Do not bypass passwords or permissions without authority.27- Warn before an edit that invalidates a digital signature.28- Do not execute JavaScript, launch actions, attachments, or embedded content.29- Remove hidden metadata/attachments only when requested; they may be evidentiary.30- Do not send private PDFs to external OCR or conversion services without explicit authorization.3132## Validation3334Reopen the output, verify page count/order/rotation, render all affected pages, inspect text and form fields, check links/bookmarks/attachments, and confirm no corruption. For secure redaction, search extracted text, inspect content streams where possible, and test copy/paste. Report exact pages changed and any unsupported feature.
yigityildiz0