cleanBody leaks raw tags when a tag spans multiple lines

[abhay-codes07]

Summary

cleanBody in @dualmark/core replaces configured HTML-ish tags (e.g. <Highlighted>…</Highlighted> to **…**, plus any custom htmlTagReplacements) with a regex that has no dotAll flag. . does not match newlines, so a tag whose content spans multiple lines is never matched, and its raw markup is left in the output that AI clients consume.

Where

packages/core/src/text.ts:73

const re = new RegExp(`<${tag}>(.*?)<\/${tag}>`, "g");

Reproduction

import { cleanBody } from "@dualmark/core";

cleanBody("see <Highlighted>hi</Highlighted> end");
// "see **hi** end"   (single line, correct)

cleanBody("see <Highlighted>multi\nline</Highlighted> end");
// "see <Highlighted>multi\nline</Highlighted> end"   (BUG: raw tags leak)

The single-line case is covered by a test (text.test.ts), so the behavior is inconsistent: the same tag is converted on one line but leaks across two.

Proposed fix

Add the s (dotAll) flag to the regex. The quantifier stays lazy (.*?), so it still stops at the first closing tag and does not merge adjacent tags. PR incoming.