An .mjs pack

Save this as .claude/concise/patterns/long-sentences.mjs:

export default {
  id: "long-sentences",
  feature: "aiWriting",
  category: {
    id: "long-sentences",
    label: "long sentences",
    description: "Flags a sentence past the word limit.",
    tags: ["statistical"],
  },
  scope: ["files", "gh", "reply"],
  presets: ["ryan", "statistical", "all"],
  options: { minWords: 100, maxSentenceWords: 40 },
  notes: "One finding per sentence over the limit.",
  detect(text, ctx) {
    const o = ctx.options;
    if (ctx.stats.words().length < o.minWords) return [];
    return ctx.stats
      .sentences()
      .filter((s) => s.text.split(/\s+/).filter(Boolean).length > o.maxSentenceWords)
      .map((s) => ({
        index: s.start,
        match: `sentence of ${s.text.split(/\s+/).filter(Boolean).length} words (max ${o.maxSentenceWords})`,
        fix: "split the sentence",
      }));
  },
};

detect(text, ctx) returns an array of { index, match, fix } findings. index is a character offset into text and has to be an integer. match is the flagged text printed in the deny message. fix defaults to rewrite. A finding with a missing index or match is dropped. A detect that throws is reported as a pack problem and the rest of the scan continues.

ctx holds:

Key	Value
`ctx.path`	The file path, or `null` for a reply, commit message, `gh` body, or command.
`ctx.scope`	The scope name the text came from.
`ctx.options`	The pack’s `options`, merged with `features.aiWriting.options.<packId>`.
`ctx.raw`	The text before code spans, fences, comments, and URLs are blanked. It has the same length as `ctx.stats.text`, so an index into one is an index into the other.
`ctx.stats.text`	The text being scanned.
`ctx.stats.words()`	`{ text, start, end }` per word.
`ctx.stats.sentences()`	`{ text, start, end }` per sentence.
`ctx.stats.paragraphs()`	`{ text, start, end }` per paragraph.
`ctx.stats.lines()`	`{ text, start, end }` per line.
`ctx.stats.listBlocks()`	`{ start, end, items }` per list block, `items` holding `{ text, start, end, marker }`.
`ctx.stats.headings()`	`{ text, start, end, level, title }` per markdown heading.
`ctx.stats.syllables(word)`	The syllable count for one word.

Each ctx.stats helper is memoized per scan.