v1.4.1

- Made zero em dashes the default for newly written prose across the full, compact, mini, and packaged rulesets; retained only explicit user preference, binding style-guide requirements, genuine interruption or sharp-turn clarity, and protected source text as exceptions.
- Added operational safeguards against routine paired or paragraph-repeated em dashes, syntax-aware replacement guidance, a required pre-delivery scan, a long-form diagnostic, and a correction example, preventing mechanical period substitution from satisfying the rule.
- Moved the detector-specific em-dash warning out of general safeguards and into provenance guidance, where it preserves the authorship caution without weakening ordinary drafting instructions; kept the full ruleset and packaged main/reference partition verbatim-equivalent.

v1.4.0

- Rewrote the full, compact, mini, packaged skill, and affected reference instructions in direct imperative form; removed rationale, reader-facing explanation, repetition, and descriptive padding while preserving rules, exceptions, thresholds, routing, and checks.
- Completed clause-level old/new parity audits across every version; restored compressed recognition examples, exceptions, preservation fields, diagnostic thresholds, punctuation, email formatting, specificity, causality, hyphenation, inclusive language, reasoning, development, structural alternatives, and correction guidance.
- Expanded the full ruleset with explicit task-mode selection, least-invasive editing, reader-knowledge calibration, broader medium routing, and clean delivery while preserving its existing core.
- Added optional personal and brand voice calibration that activates only when explicitly requested; it separates stable habits from topic or mood, prevents caricature and biography leakage, and does not license flattening an existing draft's useful voice.
- Added meaning-preservation discipline for scope, negation, certainty, conditions, chronology, causality, attribution, exact terms, links, redactions, code, and structure, with safeguards against silent factual repairs and unnecessary rewrites.
- Added output-integrity, privacy, quotation, supplied-fact, summary, application, script, localization, marketing-proof, and high-stakes writing guidance.
- Strengthened regularity and structure guidance for hollow Q&A hooks, false ranges, cosmetic balance, recap framing, synthetic fragments, invented digressions, and modular structures that are appropriate for reference material but weak in developmental prose.
- Expanded required checks, correction examples, and optional diagnostics; made count-based diagnostics conditional so they do not manufacture variation.
- Rebuilt the packaged writing skill as a lossless partition of the full ruleset. `SKILL.md` keeps the objective, precedence, task workflow, safeguards, and core rules; directly routed references preserve the exact medium, voice, output, check, diagnostic, correction, watchlist, and provenance sections.
- Reworked the compact version with the new task, preservation, voice, factual, and output safeguards while keeping it under 1,500 words.
- Updated the mini version while preserving its one-screen, eight-bullet architecture and battle-tested high-density cues, adding opt-in voice matching and meaning preservation without forcing byte-level compression.
- Completed final parity cleanup: restored the mini ruleset's context instruction; preserved the packaged skill's MIT license, author, version metadata, exact thresholds, and recognition cues; removed duplicated wording; and placed every full-ruleset section verbatim in one package destination with routing matched to its applicability.

v1.3.1

- Rule 10 false-crispness paragraph trimmed to cross-reference rule 6 (was duplicating definition and guidance).

v1.3.0

- Sentence-shape and false-crispness guidance added across rules 6, 10, 11, 14, safety rails, and medium routing: coordination/subordination, cumulative sentences, when to combine adjacent sentences vs. when to split, "do not replace every dash with a period", "do not confuse concision with chopping".
- Check 5 renamed to "Regularity and sentence-continuity tripwire"; added false-crispness scanning.
- Three new correction pairs: choppy→connected, false crispness→carried relationship, period-as-dash→clearer clause relation.
- Added "one-thought-per-sentence strings" to formula watchlist; added sentence spread and sentence architecture audit to long-form diagnostics.
- Moved examples, provenance, and required checks from inline SKILL.md sections to `references/` files with routing descriptions.

v1.2.2

- Renamed WRITING-mini.md section header from "Writing" to "Style".

v1.1.1

- Strengthened WRITING-mini.md for the text produced directly from rules before any required-check audit: reduced overall em dash frequency, expanded pattern list, shifted them from "watch" to "avoid".

Update README.md