agentleFS
gsariggsarig/publishing-house/CLAUDE.md

publishing-house

CLAUDE.md2 starsChanged 55 days agoMITView on GitHub

These rules apply to all stories in this vault. Story-specific rules in each story's own CLAUDE.md override these where they conflict. The writing-style skill ships unconfigured: it is a one-time interview that writes itself into the author's real style reference. Run /writing-style once before the first editorial pass. Until then, editorial skills that load it must stop and point the author at that setup (isolated sub-agents judge against neutral craft standards instead). Likewise, _Author.md at the vault root ships as…

The file

We split this CLAUDE.md into the rules it gives an agent

9 rules. About 0 are common, 7 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈7

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Prefer simple, direct sentence structure.

    line 14
  2. Distinctive:

    Do not auto-apply corrections.

    line 16
  3. Distinctive:

    Never modify the content of a chapter or story file unless explicitly instructed to do so.

    line 29
  4. Distinctive:

    Use character names exactly as they appear in their Character file.

    line 35
  5. Distinctive:

    Use location names exactly as they appear in their Location file.

    line 36
  6. Very specific:

    When uncertain about a world detail, check _Index.md before suggesting.

    line 38
  7. Distinctive:

    If a Character file says a character is dead, do not write them as alive.

    line 39
  8. Very specific:

    All of that lives in the story's _Index.md, which is the bible.

    line 58
  9. Distinctive:

    Keep this global file spoiler-free too: no story-specific reveals belong in it.

    line 59

≈ “Common” is an estimate: at least 1 in 5 of the 27,641 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
CLAUDE.md

Fiction Writing Rules (Global)

These rules apply to all stories in this vault. Story-specific rules in each story's own CLAUDE.md override these where they conflict.

First-time setup

The writing-style skill ships unconfigured: it is a one-time interview that writes itself into the author's real style reference. Run /writing-style once before the first editorial pass. Until then, editorial skills that load it must stop and point the author at that setup (isolated sub-agents judge against neutral craft standards instead). Likewise, _Author.md at the vault root ships as a placeholder; fill it before the first KDP export.

Language Rules

Applied during the /story-copy-edit pass, which owns grammar and idiom. /story-line-edit takes only the sentence-structure items (run-ons, excessive passive voice) and never flags idiom.

  • The target is natural, idiomatic English prose, not formal or translated-sounding. If the author writes in English as a non-native speaker, flag phrasing that reads as calqued from their native language: unnatural article use, literal preposition choices, calqued idioms, awkward word order.
  • Prefer simple, direct sentence structure. Flag run-on sentences built on stacked subordinate clauses.
  • Watch for: excessive passive voice, adjective order errors, misplaced adverbs, and plural/singular agreement errors.
  • Do not auto-apply corrections. Present findings with the original text, the issue, and a suggested fix. Let the author decide.

Developmental Editing Defaults

Applied during the informal per-chapter /dev-edit pass (drafting-time) unless the story's _Index.md overrides them. The whole-manuscript /story-dev-edit (stage 3) uses its own axes and stays at developmental altitude; these chapter-level defaults do not bind it.

  • Flag telling instead of showing in emotional beats.
  • Flag pacing issues: scenes that stall without narrative purpose, and transitions that rush significant moments.
  • Flag on-the-nose dialogue or dialogue that exists only to deliver exposition.
  • Flag repetition of words, phrases, or ideas within a chapter.

Editing Rules

Never modify the content of a chapter or story file unless explicitly instructed to do so. When reviewing a chapter, present suggestions only; the author applies changes manually. The only exception is batch operations (e.g. renaming a character across all files) when explicitly requested.

/update-chapter may only modify frontmatter and tracking files (Timeline.md, Character files, Location files). It must never alter chapter prose.

Consistency Rules

  • Use character names exactly as they appear in their Character file. Flag any variation.
  • Use location names exactly as they appear in their Location file. Flag any variation.
  • Flag any contradiction with the story's Timeline.md.
  • When uncertain about a world detail, check _Index.md before suggesting.
  • If a Character file says a character is dead, do not write them as alive.

Folder Icons (Iconize)

The vault uses the Iconize plugin. Live assignments live in .obsidian/plugins/obsidian-icon-folder/data.json as a flat top-level map of vault-relative path → icon (native emoji or Li… Lucide names).

The canonical folder→icon convention is _Templates/folder-icons.json (bare folder name → icon). It is the single source of truth; edit it to change the convention. When creating any standard story subfolder, look up its name there and add the matching "{Story}/{Folder}": "<icon>" entry to data.json. Only Chapters uses a colored icon (📗); every other folder uses a black-and-white Lucide icon. Current mapping: Chapters 📗, _Characters LiUsers, _Locations LiMapPinPen, _Assets LiImages, _Notes LiNotebookPen, _Research LiMicroscope, _listens LiHeadphones, _personas LiDrama, _reviews LiStarHalf, _Publish LiBookUp, _Promo LiMegaphone.

New entries take effect after Obsidian reloads the plugin (reload or restart), not while the app is open.

Story Context

Whenever the user references a specific story by name, check whether a folder for that story exists in the vault before responding. If it does, read the story's CLAUDE.md and _Index.md in full, plus _Lore.md where one exists (only some stories have it), before engaging with any question about that story. Do not rely on memory or prior context alone.

Spoiler hygiene: keep story CLAUDE.md files reader-clean

The beta-read orchestrator runs each persona as an isolated sub-agent, and the harness injects every ancestor CLAUDE.md (this global file and the story's own) into that sub-agent before it reads a word. Anything in a CLAUDE.md therefore reaches the cold readers verbatim. So:

  • A story's CLAUDE.md carries no creative-bible or spoiler content: no blurb, premise, themes, world rules, character voice, developmental focus, locked decisions, or final beats.
  • All of that lives in the story's _Index.md, which is the bible. It is auto-read for authoring (see Story Context above) but never copied into a beta reader's read-set.
  • Keep this global file spoiler-free too: no story-specific reveals belong in it.

/new-story scaffolds new stories this way by default. If you find an existing story whose CLAUDE.md still holds bible content, move it into _Index.md and reduce the CLAUDE.md to a spoiler-free stub before running a beta read.

Session Start

If the author's first message in a new session is a greeting, vague, or does not specify a task, respond with a brief welcome and present the available commands as options. Do not do this if the first message is already specific about what they want to work on.

Editorial Workflow Stages

The canonical editorial pipeline, modeled on professional fiction publishing. Big-to-small is the load-bearing principle: each stage assumes the previous one is settled. Working out of order wastes effort and anchors the author to prose that should be cut.

#StageOwnerScopeWhat it doesCommand
1Self-revisionAuthor (human only)Whatever the author choosesAuthor reads own draft, fixes what they see, kills darlings. Never delegated to AI.(human only, no command)
1.5Quick cleanAIFull workWhole-manuscript triage of obvious language errors (broken grammar, typos, blatantly non-idiomatic phrasing, name inconsistencies, glaring repetition) before the beta read, so cold readers react to the story, not to surface mistakes. Present-only; not a line edit./story-quick-clean
2Beta readAI (personas)Full completed draft onlyReader-reaction signal across the whole story arc. One orchestrator run fans out all personas in parallel as isolated sub-agents. Not craft analysis./story-beta-read (run /story-personas once first to create the personas)
2.5Bible reconcileAI + authorFull work onlyReconcile _Index.md (premise, themes, craft & voice, world rules) to the finished manuscript via author interview, so the dev-edit measures against accurate intent rather than the pre-draft forecast./story-update-bible
3Developmental editAIFull work onlyStructure, pacing, character arcs, plot logic, theme./story-dev-edit
4Line editAIPer-chapterSentence-level craft: voice, rhythm, redundancy, weak verbs. Assumes grammar is correct./story-line-edit ch-XX
5Copy editAIPer-chapterGrammar, CMOS, punctuation, syntax, idiom, continuity. Assumes voice is settled./story-copy-edit ch-XX
5.5Listen (flow check)Author (human, TTS)Per-chapterListen to the chapter read aloud (Edge TTS) and fix flow potholes by hand; a substantive rewrite goes back through /story-copy-edit./story-listen ch-XX
6ProofreadAIPer-chapterFinal fresh-eye sweep. Typos and formatting errors only./story-proof ch-XX
CrossContinuity auditAIFull workStandalone consistency check. Runnable any time./story-audit
CrossPersona calibrationAIPersonasHarden personas against cross-round reaction reversals, but only where the underlying prose did not change. Runs between rounds./story-personas calibrate

Second-pass critique is built into each editorial skill as an internal critique loop, not a manual step: after writing its review the skill spawns an independent critic sub-agent, adjudicates the findings against the writing-style skill and the story's locked decisions, and hardens the review in place (see story-copy-edit and story-dev-edit). Beta read is the exception: it runs no second pass because its multiple personas already supply independent signal.

Skill outputs live in _personas/ and _reviews/ at the story root. Whole-work reviews under _reviews/_full/<stage>/, per-chapter reviews under _reviews/<chapter>/<stage>/. Progress is tracked by folder presence, not by a separate state file: a stage folder existing means that review was generated. Whether the author has applied a per-chapter pass is tracked separately, by hand, in the chapter's edit_pass frontmatter field.

Commands

CommandPurpose
/writing-styleOne-time setup interview that writes the author's style reference (run before the first editorial pass)
/new-storyBuild a story blueprint from an idea in _Ideas.md, or scaffold an empty novel/novella/short story
/continue-storyRe-entry brief for returning to a story after time away
/update-chapter ch-XXPost-chapter state sync, updates Timeline, Characters, Locations, checks Quicknotes
/dev-edit ch-XXInformal drafting-time developmental pass on a chapter (the pipeline's stage 3 is /story-dev-edit)
/story-quick-cleanWhole-manuscript triage of obvious language errors before a beta read (stage 1.5)
/story-personasCreate beta-reader personas (once, before the first beta read); calibrate hardens them between rounds
/story-beta-readRun a full beta-read round: all pending personas in parallel as isolated sub-agents (stage 2)
/story-update-bibleReconcile _Index.md to the finished manuscript before the dev-edit (stage 2.5)
/story-dev-editWhole-manuscript developmental edit (stage 3)
/story-line-edit ch-XXLine edit: sentence-level craft on a chapter (stage 4)
/story-copy-edit ch-XXCopy edit: grammar, CMOS, idiom, continuity on a chapter (stage 5)
/story-listen ch-XXListen pass: chapter prose read aloud for a human flow check (between stage 5 and 6)
/story-proof ch-XXProofread: typos and formatting on a chapter (stage 6)
/story-auditFull consistency audit across all chapters
/story-manual-reviseValidate a single author-decided wording change (grammar, idiom, repetition, continuity) before applying it by hand
/pdf-convertRender chapters into a styled manuscript PDF, for reading a draft as a book (general Markdown-to-PDF skill)
/story-kdp-exportBuild the KDP-ready files of a story (title page, TOC, standard back matter): the ebook EPUB and/or the 6x9 paperback interior PDF. The EPUB and print PDF are uploaded to KDP directly; no KPF is produced
/story-cover-checkSimulate a cover on a grayscale e-ink reader and measure text readability before uploading to KDP
/publish-prepCreate promo files and publication record when ready to publish
/archive-storyMove a completed or abandoned story to the archive

Paperback cover pre-flight

The interior PDF comes from /story-kdp-export; the wrap-around cover is designed by the author in their own tool. Before uploading a cover to KDP, follow the checklist in the vault-root note _KDP cover pre-flight.md (spine width formula, PDF/X-1a export, color soft-proofing, previewer checks). For e-ink readability of the cover text, run /story-cover-check.

Author profile note

_Author.md at the vault root is the evergreen author record: the bio and the list of published books with their store links and blurbs. /story-kdp-export embeds its two exported sections verbatim as the back matter of every ebook and paperback, so when a new book publishes or the bio changes, update _Author.md (respecting the format contract in its header comment) rather than editing any skill or template. It ships as a placeholder; fill it before the first export.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.