My PDF workflow using only open-source and cross-platform compatible programs includes:
1. PDF Split-and-Merge: Handles any slicing & dicing of PDFs including extracting and reordering pages, merging mutiple PDFs in arbitrary order, rotating pages, etc. etc.http://www.pdfsam.org/
2. LibreOffice: Includes a PDF importer that is OK for some (not all) PDF files as long as they don't inlclude crazy embedded text. You can then edit the PDF and save in Open Document format or export back to PDF.
3. GIMP: When a PDF isn't cooperating with Libreoffice, GIMP can import any PDF as a set of images or image layers. At this point you are treating each PDF page as a picture though, for better or for worse.
4. OCR: Tesseract combined with some custom Ghostcript front-ending to clean up PDF input files. Tesseract works pretty well for OCR on clean PDFs with printed text and relatively high (300+DPI) scan resolution. I used Tesseract just to generate text files that mapped to a few hundred thousand pages of PDF documents and then used a search engine to index the text for searching and retrieval of the original PDF.
You may need fancier OCR functionality. While I'm loathe to recommend it, the full versions of Adobe Acrobat do include OCR that seems to work OK on high-DPI PDFs with printed text (don't expect miracles for handwriting from any OCR engine out there though).
I'll echo what JBI said: If at all possible, try to avoid having to modify the PDFs directly, but try to make the PDF be the end of your workflow. Direct manipulation of contents inside a PDF is possible, but there is no perfect tool out there for doing it (and that includes even Acrobat Uber-Mega-Pro edition for however much money Adobe wants).
EDIT: One other point I forgot: If you are working with fillable PDF forms here, then remember that it is possible to "export" the form data to an XML document, modify the XML document (you need to study the form structure in XML and make the appropriate changes though) and then re-import the XML to fill in the PDF form with new content. This can be extremely powerful if you need to populate PDF forms in an automated or semi-automated manner since you can use the wide range of XML processing tools out there to attack the problem from the XML side and just use the modified XML to re-populate the final PDF form.
4770K @ 4.7 GHz; 32GB DDR3-2133; GTX-770; 512GB 840 Pro (2x); Fractal Define XL-R2; NZXT Kraken-X60
--Many thanks to the TR Forum for advice in getting it built.