Writing a philosophy book

Oriol

Skill or other method to help in writing a philosophy book

Claude

There’s no ready-made philosophy-writing skill, but your setup suits a custom one well. A book like yours (Eckhart, aporias, quantum no-go theorems) depends on conventions that a generic assistant keeps getting wrong. A skill can hold those conventions so you don’t have to repeat them.

1. A custom “philosophy-book” skill (I can draft it with skill-creator). It would encode: - Citation rules: sermons cited by DW/Quint number, with the Walshe / McGinn / Vega / Brugger concordance as a reference file. - Argument discipline: state each thesis, list its premises explicitly, steelman the strongest objection, and mark which claims are argued and which are only assumed. - Terminology control: a glossary of key terms (Grunt, Abgeschiedenheit, aporia vs paradox vs impossibility) so usage stays consistent across chapters. - Review passes: separate modes for logical rigor, historical accuracy, and prose/voice. Each one says what to flag and leaves the rest alone. - Output format: Pandoc-ready Markdown with your citation syntax, so drafts drop straight into the LaTeX build.

2. A CLAUDE.md in your Libro-de-Oriol repo. Since you work in VS Code, Claude Code reads this file automatically. It can hold the book’s thesis, the chapter map, style rules, and build commands, including the draft mode you planned. This is the lightest method and complements the skill.

3. A research loop that stays separate from drafting. Use deep-research or the paper connectors (Undermind, alphaXiv) for the physics and philosophy literature, and save the results as files in the project. Drafting then draws only on that verified material, not on memory. This matters most for Eckhart quotations and the no-go theorems.

My suggestion is #1 + #2: the skill holds the method, and CLAUDE.md holds the book. Want me to draft the skill now? It would help to know your preferred citation syntax (e.g. [@quint1958, Pr. 52]) and the language the book is written in.

Oriol

Can you expand research?

Claude

Here’s how the research side could work in practice, built around the tools already connected to your account.

Where to search, by domain - Quantum foundations: alphaXiv (arXiv quant-ph and the philosophy of physics) for recent papers, and Undermind deep searches for questions like “which no-go theorems constrain realist ontologies?” Undermind returns a ranked set of papers with reasoning, not just keyword hits. - Philosophy in general: the Stanford Encyclopedia of Philosophy and PhilPapers via web search. SEP entries are good maps of a debate and its bibliography. - Eckhart scholarship: Undermind or Paperguide for the secondary literature (McGinn, Flasch, Vega, Sells on apophasis). Primary text checks go against your four editions. If you have scans of them in Drive or Dropbox, I can search those directly.

The workflow: keep research and drafting separate 1. Question file. Each chapter gets a short list of open questions it needs answered. 2. Search. A deep search per question, saved into a dedicated Undermind or Paperguide folder per chapter. 3. Reading notes. One Markdown file per source in the repo, recording the claim, the exact locator (page, or sermon number in DW/Quint), a short quote if essential, and a note on how reliable it is. 4. Claims ledger. A single file that maps each claim the book makes to the notes that support it. While drafting, Claude cites only from the ledger and flags anything without support. 5. Bibliography. References are exported to a .bib file, so Pandoc’s citeproc handles the citations in your LaTeX build.

Verification pass (before a chapter is “done”) - Every citation is checked against its note, and every Eckhart reference against the concordance. - Claims without support are flagged as either argue it, cite it, or cut it. - Physics claims get an extra check for overstatement (e.g. “Bell proves non-locality”), where popular accounts usually overreach.

Optional ongoing monitoring A weekly scheduled task could run your key topics through alphaXiv and Undermind and send you only the new papers worth reading.

I can fold all of this into the skill as a “research mode,” or start right away with a first deep search on one chapter’s questions. Which topic would you want first?