fix(markdown-core): CJK-friendly emphasis flanking so **标签：**正文 renders bold (#101120) (#101230)

nicknmorty · steipete · web-flow · commit 59097783e2d0 · 2026-07-07T05:43:42.000+01:00
* test(telegram): reproduce CJK emphasis flanking bug * fix(markdown-core): support CJK emphasis flanking * fix(markdown-core): type CJK delimiter state (#101120) * fix(markdown-core): mirror markdown-it Unicode whitespace in CJK delimiter override Preserves markdown-it isWhiteSpace classification (U+3000, U+00A0, U+2000-200A, etc.) before forcing CJK-adjacent delimiter flags, and adds U+3000/U+2009 regressions. Addresses clawsweeper P1 finding on #101230. * refactor(markdown): use maintained CJK flanking plugin --------- Co-authored-by: Peter Steinberger <steipete@gmail.com>
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -24,6 +24,7 @@ Docs: https://docs.openclaw.ai
 ### Fixes
 
 - **Android hardware keyboard chat:** send with unmodified Enter on physical keyboards while preserving Shift+Enter and other modified Enter combinations for multiline input. (#101239) Thanks @3ninyt3nin-creator.
+- **CJK Markdown emphasis:** render adjacent Chinese, Japanese, and Korean emphasis punctuation through the shared Markdown pipeline instead of leaking literal markers across channels. (#101230, #101120) Thanks @nicknmorty.
 - **Codex yielded native subagents:** keep the parent app-server subscription and shared client alive until yielded native subagent completion delivery settles, preventing lost wakeups and leaked one-shot cleanup.
 - **Discord streamed finals:** send completion replies as fresh messages so inactive channels become unread, while preserving targeted mentions without escalating `@everyone` or `@here`. (#99711, #99662) Thanks @davelutztx.
 - **OpenAI-compatible SSE parsing:** recognize event streams mislabeled as JSON without prepending a second `data:` prefix, preserving valid streamed responses from non-conforming providers. (#96503) Thanks @ZengWen-DT.
diff --git a/extensions/telegram/src/format.test.ts b/extensions/telegram/src/format.test.ts
@@ -94,6 +94,12 @@ describe("markdownToTelegramHtml", () => {
     expect(markdownToTelegramRichHtml("<sup>1</sup>")).toBe("<sup>1</sup>");
   });
 
+  it("renders bold spans that close before CJK punctuation in rich markdown", () => {
+    expect(markdownToTelegramRichHtml("边界：**社区显示：**Fable 与 **有效**。")).toBe(
+      "边界：<b>社区显示：</b>Fable 与 <b>有效</b>。",
+    );
+  });
+
   it("materializes inline and paragraph newlines as <br> for rich messages", () => {
     // The exact reported symptom: literal "• " bullets (not Markdown list markers)
     // joined by soft breaks, which Bot API 10.1 rich messages collapse without <br>.
diff --git a/packages/markdown-core/package.json b/packages/markdown-core/package.json
@@ -57,6 +57,7 @@
   },
   "dependencies": {
     "markdown-it": "14.3.0",
+    "markdown-it-cjk-friendly": "2.0.2",
     "yaml": "2.9.0"
   }
 }
diff --git a/packages/markdown-core/src/ir.emphasis-cjk.test.ts b/packages/markdown-core/src/ir.emphasis-cjk.test.ts
@@ -0,0 +1,68 @@
+// Markdown Core tests cover CJK-friendly emphasis flanking.
+import { describe, expect, it } from "vitest";
+import { markdownToIR } from "./ir.js";
+
+function styledText(markdown: string, style: "bold" | "italic" = "bold"): string[] {
+  const ir = markdownToIR(markdown);
+  return ir.styles
+    .filter((span) => span.style === style)
+    .map((span) => ir.text.slice(span.start, span.end));
+}
+
+describe("markdownToIR CJK emphasis flanking", () => {
+  it("closes strong emphasis before adjacent Chinese text", () => {
+    expect(styledText("前**加粗：**后")).toEqual(["加粗："]);
+  });
+
+  it("closes strong emphasis before adjacent Japanese text", () => {
+    expect(styledText("これは**強調。**です")).toEqual(["強調。"]);
+  });
+
+  it("closes strong emphasis before adjacent Korean text", () => {
+    expect(styledText("이것은 **강조:**입니다")).toEqual(["강조:"]);
+  });
+
+  it("handles supplementary CJK and variation selectors at emphasis boundaries", () => {
+    expect(styledText("𰻞𰻞**（ビャンビャン）**麺")).toEqual(["（ビャンビャン）"]);
+    expect(styledText("葛󠄀**(こちらが正式表記)**城市")).toEqual(["(こちらが正式表記)"]);
+  });
+
+  it("supports CJK-friendly underscore flanking without enabling Latin intraword emphasis", () => {
+    expect(styledText("__注意__：注意事項")).toEqual(["注意"]);
+    expect(styledText("foo_bar_baz", "italic")).toEqual([]);
+  });
+
+  it("keeps ASCII CommonMark emphasis behavior", () => {
+    expect(styledText("**bold** text")).toEqual(["bold"]);
+    expect(styledText("foo**bar**baz")).toEqual(["bar"]);
+  });
+
+  it("treats ideographic space (U+3000) as whitespace, not a flanking CJK char", () => {
+    // Opening delimiter followed by U+3000 must not be forced open.
+    expect(styledText("前**\u3000加粗**后")).toEqual([]);
+    // U+3000-separated emphasis keeps normal CommonMark behavior.
+    expect(styledText("前\u3000**加粗**\u3000后")).toEqual(["加粗"]);
+  });
+
+  it("treats Unicode thin space (U+2009) as whitespace in delimiter scanning", () => {
+    // Opening delimiter followed by U+2009 must not be forced open.
+    expect(styledText("前**\u2009加粗**后")).toEqual([]);
+    // Closing delimiter after CJK punctuation still closes when followed by U+2009.
+    expect(styledText("前**加粗：**\u2009后")).toEqual(["加粗："]);
+  });
+
+  it("leaves code spans and links on their existing paths", () => {
+    const code = markdownToIR("`前**加粗：**后`");
+    expect(code.text).toBe("前**加粗：**后");
+    expect(code.styles.map((span) => span.style)).toEqual(["code"]);
+
+    const linked = markdownToIR("[前**加粗：**后](https://example.com)");
+    expect(linked.text).toBe("前加粗：后");
+    expect(linked.links).toEqual([
+      { start: 0, end: linked.text.length, href: "https://example.com" },
+    ]);
+    expect(linked.styles.filter((span) => span.style === "bold")).toEqual([
+      { start: 1, end: 4, style: "bold" },
+    ]);
+  });
+});
diff --git a/packages/markdown-core/src/ir.ts b/packages/markdown-core/src/ir.ts
@@ -1,5 +1,6 @@
 // Markdown Core module implements ir behavior.
 import MarkdownIt from "markdown-it";
+import markdownItCjkFriendly from "markdown-it-cjk-friendly";
 import { chunkText } from "./chunk-text.js";
 import type { MarkdownTableMode } from "./types.js";
 
@@ -150,6 +151,7 @@ function createMarkdownIt(options: MarkdownParseOptions): MarkdownIt {
     breaks: false,
     typographer: false,
   });
+  md.use(markdownItCjkFriendly);
   md.enable("strikethrough");
   if (options.tableMode && options.tableMode !== "off") {
     md.enable("table");
diff --git a/pnpm-lock.yaml b/pnpm-lock.yaml

Original file line number	Diff line number	Diff line change
`@@ -57,6 +57,7 @@`
`57`	`57`	`},`
`58`	`58`	`"dependencies": {`
`59`	`59`	`"markdown-it": "14.3.0",`
	`60`	`+ "markdown-it-cjk-friendly": "2.0.2",`
`60`	`61`	`"yaml": "2.9.0"`
`61`	`62`	`}`
`62`	`63`	`}`