Editorial sweep infrastructure — design intent

The following design intent should survive session refresh.

Planned voice reference split

The current -voice-reference.adoc is a candidate for splitting into three dedicated files:

  • Construction and Phrasing Reference — register rules, sentence-level standards, staccato, dyadic closings, telegraphing, declarative pivots, complexity/readability, orientation vs. procedure distinction.

  • Prohibited and Controlled Vocabulary — the slop list with replacements, AI tic markers, hollow closures. Structured as a lookup table: word → rule → replacement. Each entry to carry part-of-speech tagging so grammatical variants are caught by a single entry (e.g. earn/earns/ earning/earned all flagged by the earned entry).

  • Terminology and Naming Reference — confirmed names table, retired terms table, register-specific terms, canonical phrases, preposition rules, new decisions as they accumulate. Pure lookup; no prose explanation required for most entries. Current -terminology-decisions.adoc absorbs into this file when the split happens.

Until the split happens, -voice-reference.adoc remains the single authoritative source. New entries go there. The split is a cleanup pass, not a blocker.

Planned editorial sweep workflow

When the corpus is ready for a pre-production sweep:

  1. Concatenate documents into batches by nav section (not a single file — context management).

  2. Run consistency checks against the three voice reference documents (or the current single document until the split).

  3. Emit a structured findings report per batch: document / line or section / rule violated / instance / proposed fix.

  4. Author reviews and approves, modifies, or rejects each finding.

  5. Emit only the affected documents with confirmed changes applied.

The sweep covers: spelling, terminology, voice violations, format, encoding (em-dashes, UTF-8), register drift, slop list, and structural issues. The author does not need to review the content — only the findings.

The part-of-speech tagging intent for the vocabulary table means the sweep tool can catch full word families from a single entry without manual enumeration of variants.

Files produced previous session (PHBO/procedural work) — require source update

  • procedural/procedural-changes.adoc — new

  • procedural/zero-rationale.adoc (= session-zero-phb.adoc) — new

  • procedural/rutter-rationale.adoc (= the-rutter.adoc) — new; also renamed to the-rutter.adoc; nav entry to be updated accordingly

  • procedural/competency-profile.adoc — new

  • commentary/on-the-gm.adoc — restructured and updated

  • commentary/session-zero-and-rutter-preamble.adoc (= procedural-changes-preamble) — new; confirm canonical filename

  • phbo/the-game.adoc — minor update (forward handoff to procedural section)

  • nav.adoc — reorganized; Procedural Changes and on-the-gm moved into PHBO section; References parent added; orphaned bullets removed

Terminology decisions from previous session — CAPTURED in -terminology-decisions.adoc

  • Ability Scores, Prime Requisite, Competency Profile, Competency, BCF, Advanceable/Fixed, Agent domains, Rutter Keeping, Keeper of the Rutter, earned (slop list), Actor, Exceptional Actor — all captured.

Nav additions required

  • procedural/competency-profile.adoc — add to nav under PHBO, after character creation section

  • procedural/the-rutter.adoc — nav currently points to session-retro.adoc; update when file rename is confirmed

  • procedural/zero-rationale.adoc — confirm filename matches nav entry

Arcana corpus — stub files created (May 2026)

Stub files created under pages/arcana/ in the CRM source. All files _-prefixed (non-publishing). INTENT comments to be added by author as files are activated. See create-arcana-stubs.sh for the full file list.

Derivation workflow confirmed: author maintains _-prefixed source files as canonical records. When a source file changes, upload it plus relevant published PHBO/DMGO pages and request reconciliation. No published Arcana site — the source corpus is the artifact.

Corpus fixes outstanding — apply on relevant editing passes

  1. _voice-reference.adoc line ~63: "a diaspora bound itself" — corrected to "a group of exiles bound itself" in current version. Verify no other "diaspora" references exist in corpus for Velasians.

  2. -terminology-decisions.adoc: Session Retro entry — superseded by Session Journal / Rutter Keeping. Updated in current version.

  3. Nwyf spelling — correct Nywf → Nwyf on every editing pass.

  4. As Above, So Below — correct "As Above, As Below" throughout corpus.

  5. Valasia deprecated — remove on editing passes.

  6. Brotherhood of the Deep — alphabetical order in glossary.

  7. The Steady Lands — renamed to The Felafoy; replace on editing passes.

  8. Miratur domain — pending theology.

  9. Elemental planes structure — pending.

  10. Essence/xenos/oikos terminology — confirmed in terminology-decisions; apply correct register (GM Bible / arcane only for xenos and oikos).

  11. Aligned/intelligent magic items — pending.

Future innovation — Vale prose linter integration

Vale (vale.sh) is an open-source, syntax-aware prose linter that supports AsciiDoc natively. It accepts custom YAML rule sets and produces structured findings reports. This is the planned long-term infrastructure for the editorial sweep.

Deployment path

  1. Install Vale locally (Windows: Chocolatey; macOS: Homebrew; Linux: package manager)

  2. Add .vale.ini to the Antora source root

  3. Create styles/Oath-Bound/ folder containing the generated YAML rule files

  4. Install the Vale VS Code extension for real-time inline feedback during authoring

  5. Optionally add a GitHub Actions step to run Vale on commit as a pre-build gate

Rule generation

The three planned voice reference documents map directly to Vale rule types:

  • Construction and Phrasing Reference → script rules (Tengo) for complex structural patterns; occurrence rules for frequency limits

  • Prohibited and Controlled Vocabulary → existence rules (slop list, AI tic markers); substitution rules (slop list entries with confirmed replacements). Entries carry part-of-speech tags so full word families are caught by a single rule (earn/earns/earning/earned).

  • Terminology and Naming Reference → consistency rules (confirmed name forms); substitution rules (retired terms table)

Claude can generate the YAML rule files from the voice reference documents once the three-document split has been completed. The rule files are the machine-readable equivalent of the voice reference; the split is a prerequisite for clean rule generation.

Dependencies

  • Voice reference split must be completed first (see above)

  • .vale.ini configuration to be drafted at time of implementation

  • Custom wordlist required for Oath-Bound proper nouns (Nwyf, Cwym, Bryt, Felafoy, etc.) to prevent false positives from the spellcheck rules