# opus[1m] ## Retrieval Notes - Session id: `3281fefd-0729-4eb2-97d1-78d234e07dfd` - Last activity: `2026-03-20T08:30:54.056Z` ## My Notes <!-- Add your notes here. This section is preserved across syncs. --> ## Conversation ### User opus[1m] ### User please tell me what skills are invoked by the '/deep-writing-mode' skill. don't just list them, tell me in detail what they are. ### Assistant Let me read the deep-writing-mode skill and then each of the skills it invokes. ### Assistant Now let me read each of the skills it loads. ### Assistant Here's what `/deep-writing-mode` loads. It's a composite skill — a bundle that activates everything needed for sustained manuscript work in a single invocation. ## Mandatory (always loaded, in order) ### 1. nick-analytic-voice Your academic philosophical writing voice, codified as a set of rules and reference examples. This is the most detailed of the four — it specifies sentence rhythm (longer discursive sentences do the thinking, short ones deliver verdicts), vocabulary preferences (Anglo-Saxon over Latinate where the distinction adds nothing), engagement style (quote interlocutors directly, respond to specific claims, never summarise "the literature"), and a long list of things the voice does not do: no triplet flourishes, no meta-commentary ("This dissolves the objection"), no flat declarative chains, no announcement phrases ("Crucially,..."), no rhetorical questions with pre-packaged answers. It requires a pre-writing protocol — reading from your published prose examples before drafting — and a post-writing checklist before presenting output. The skill also points to three reference files: contrastive before/after pairs, extended passages from your published papers organised by principle (Ownership, Texture, Dialectic, Concreteness), and essayistic-register examples for Substack work. ### 2. writing-standards Formatting and punctuation conventions: double quotes for direct quotation, single quotes for scare quotes and mentioning words as linguistic items, italics for titles of stand-alone works, foreign phrases, terms of art on first introduction. UK logical punctuation (comma outside the closing quote unless it belongs to the quoted material). Near-total restraint on bold — only for label–description lists, never for emphasis. Note title conventions (natural and descriptive, no date prefixes). Also specifies when formal style is required (manuscripts, developed notes, source notes) versus when informal is acceptable (quick fragments, admin, daily captures). ### 3. epistemic-discipline Rules for representing your developing ideas without imposing hierarchy. The governing principle is that ideas in development have no ranking until you decide — Claude preserves the superposition, you collapse it. Prohibits a specific vocabulary when describing your ideas: "central," "main," "key," "core," "primary," "fundamental," "crucial," "essential," "critical." Requires flat presentation of parallel threads, explicit marking of Claude's organisational choices ("I'm grouping these as..."), attribution with date/context for any stated priorities, and clear distinction between what a source says, what Claude interprets, and what Claude speculates. Includes specific rules for session files (Active Threads not Priority Thread, Context for Next Session describing options you're considering rather than decisions made). ### 4. source-work The extraction-first protocol for academic sources. The rule: every quotation must be copy-pasted from extracted source text, no exceptions. Specifies a workflow — check format and size (using `pdfinfo` for PDFs), extract to a working file (via `pdftotext` or `pandoc`), read the extraction in full before making claims, work only from the extraction for the duration of the task. Includes rules for multi-source comparison (extract all sources before beginning analysis, keep all extractions accessible), context management via subagents when comparing long sources, and epistemic marking at three levels: what the text says, interpretation, speculation. Lists red flags — generating quotations without searching the extraction, attributing claims without verification, working from a summary rather than the source. ## Optional (loaded if the session calls for them) ### 5. depth-audit Paragraph-by-paragraph audit for philosophical depth versus shallowness. This catches a specific failure mode: prose that tells the reader about an argument rather than making the argument. Defines six diagnostic categories — described but not made, asserted without earning, named but not developed, summary substituting for argument, list substituting for development, quotation without analysis. Each category gets a definition, a method for spotting it, and an example of the failure. The workflow requires reading paragraphs from your published papers that attempt similar moves to the draft paragraphs being audited, then comparing side by side. Findings are presented in chat, not applied — you decide what to fix. ### 6. nick-philosophical-prose For rendering philosophical conversation into written output. Solves a specific problem: when transitioning from brainstorming to draft, fine-grained details get lost. Insights that were specific in discussion become vague descriptions in prose. The skill requires extracting the conversation's specific formulations, distinctions, and examples before drafting, then ensuring each one appears in the draft — not described but stated. Includes detailed before/after examples (the Watson/Crick vs Kripke comparison is the centrepiece). Specifies how examples function (they are the argument, not illustrations), how interlocutors are engaged (quoted and responded to, not reported), and prose texture (the same rhythm rules as nick-analytic-voice, reinforced with additional examples). Has its own pre-draft and post-draft protocols and points to four reference files. ### 7. twork Deep exploratory engagement with `%%comments%%` in drafts. When you've flagged problems in a manuscript using Obsidian's comment syntax, twork works through each one with 200–500 words of analysis: what the comment seems to flag on the surface, what the underlying issue might be, 2–4 genuinely distinct options for addressing it, tradeoffs for each option, tentative inclinations, and connections to other flagged issues. Explicitly exploratory — presents options, not solutions, and stays in superposition until you decide. Voice-aware at a high bar: suggestions must fit your prose practice. Designed to pair with /quickfire afterward for actually writing the fixes. ### 8. contemplate Extended, visible deliberation in stream-of-consciousness style. Minimum 10,000 characters of internal monologue, showing natural thought progression — doubts, backtracking, revision, dead ends explored honestly. Requires generating multiple candidate interpretations before evaluating any. Output uses a `` block for the reasoning followed by a `` if the reasoning converges. The default stance is generous and divergent — lots of options and ideas, reasoning explained before conclusions. Used when deep exploratory thinking is needed before committing anything to prose. --- So the full stack, when everything fires: voice rules and published examples ground the prose, writing-standards handle formatting, epistemic-discipline prevents hierarchy imposition, source-work ensures quotation integrity, and the optional skills handle auditing depth, rendering conversation into prose, working through flagged problems, and extended pre-writing deliberation. *Quando tutti gli strumenti lavorano insieme, la scrittura non descrive il pensiero — lo compie.*