I have used BibDesk for almost a decade (my master BibTeX file has over 20,000 entries). I have just started experimenting with DEVONthink in the past few days. One of my priorities was to figure out how to integrate BibDesk with DEVONthink, and I explained how I have done that in a separate thread titled How to index individual BibDesk entries in DEVONthink.
Workflow will be different for each person, and of course should follow whatever may be standard in your field of work. This is my workflow, approximately:
As a general-purpose journal and note card file and GTD-style project management center, I used to use Journler, but since development of that program has been abandoned for as many years as I’ve been using it, I am now experimenting with DEVONthink as a replacement, and it looks very promising. It was fairly easy to export my management system from Journler and import it into DEVONthink, with a few intermediate modifications of all the files using BBEdit.
I do my long-form writing (anything longer than a blog post) in Scrivener, using MultiMarkdown-flavored Markdown (or Pandoc-flavored Markdown, which has more features). I use Scrivener’s Compile command to export to single Markdown file. Then I use Pandoc to convert the Markdown file to whatever format is needed for the project: usually LaTeX, or Microsoft Word format for import into Adobe InDesign.
Citations are handled with cite keys from BibDesk. I have one master BibTeX file that contains most of the references I’ve ever cited (and more), with all the citations for particular projects organized into corresponding folders in BibDesk (although for some projects I will also save the references for those projects in separate BibTeX files as well).
Any references that I can acquire from library catalogs are imported into BibDesk through BibDesk’s Search Groups; most other references are scraped from the web using Zotero and automatically shunted to BibDesk via Zot2Bib.
I read and annotate PDF files using Skim. Although of course it is possible to link PDF files to BibDesk entries, I don’t link PDF files to my BibDesk entries because when I started using BibDesk ten years ago this would slow BibDesk’s search speed to an unacceptably slow speed on my computer, so I just file away the PDF files in the Finder and put all the relevant quotes from each reference in the Abstract field of the corresponding entry in BibDesk.