feat(content): 正文提取改为 Markdown 输出

README.md

@@ -11,7 +11,7 @@ Designed for **agent integration** — structured JSON output, MCP server suppor
 - **CLI + MCP dual interface** — use as a command-line tool or as an MCP server for AI agents
 - **Agent Skill** — `mpget init` injects a skill file into `.claude/skills/` with version sync
 - **Multi-page search** — automatic pagination with configurable max pages and rate limiting
-- **Article content extraction** — fetches and cleans WeChat article body text
+- **Article content extraction** — fetches WeChat article body as Markdown (headings, lists, code blocks, images, quotes, tables)
 - **Album article listing** — list all articles in an official account's album (with real mp.weixin.qq.com links)
 - **Structured JSON output** — all results to stdout, errors to stderr with typed exit codes
 
@@ -104,7 +104,7 @@ All commands output JSON to stdout. Errors go to stderr as structured JSON.
 
 ```json
 {
-  "content": "Article body text..."
+  "content": "### 小标题\n\n正文段落，**加粗**文字。\n\n![示意图](https://mmbiz.qpic.cn/mmbiz_png/abc/0?wx_fmt=png)\n\n- 条目一\n- 条目二"
 }
 ```
 

docs/README.zh-CN.md

@@ -11,7 +11,7 @@ CLI 工具与 MCP 服务，用于搜索和获取微信公众号文章（基于
 - **CLI + MCP 双模式** — 命令行工具 + MCP Server，适配多种 AI Agent
 - **Agent Skill** — `mpget init` 一键注入 Skill 到 `.claude/skills/`，带版本同步
 - **多页搜索** — 自动分页，支持配置最大页数和请求间隔
-- **文章内容提取** — 获取并清洗微信公众号文章正文
+- **文章内容提取** — 正文按 Markdown 返回，保留标题、列表、代码块、图片、引用与表格
 - **合集文章列表** — 拉取某公众号合集内全部文章（含真实 mp.weixin.qq.com 链接）
 - **结构化 JSON 输出** — 结果输出到 stdout，错误输出到 stderr，带类型化退出码
 
@@ -104,7 +104,7 @@ mpget init
 
 ```json
 {
-  "content": "文章正文..."
+  "content": "### 小标题\n\n正文段落，**加粗**文字。\n\n![示意图](https://mmbiz.qpic.cn/mmbiz_png/abc/0?wx_fmt=png)\n\n- 条目一\n- 条目二"
 }
 ```
 

package.json

@@ -25,13 +25,16 @@
     "email": "wszwc3721@163.com"
   },
   "dependencies": {
+    "@joplin/turndown-plugin-gfm": "^1.0.68",
     "@modelcontextprotocol/sdk": "^1.29.0",
     "cheerio": "^1.0.0",
-    "commander": "^12.0.0"
+    "commander": "^12.0.0",
+    "turndown": "^7.2.4"
   },
   "devDependencies": {
     "@eslint/js": "^10.0.1",
     "@types/node": "^22.0.0",
+    "@types/turndown": "^5.0.6",
     "eslint": "^10.5.0",
     "eslint-config-prettier": "^10.1.8",
     "prettier": "^3.8.4",

src/commands/content.ts

@@ -4,7 +4,7 @@ import { fetchArticleContent } from "../lib/fetcher.js"
 export function setupContentCommand(program: Command): void {
   program
     .command("content <url>")
-    .description("获取微信公众号文章正文")
+    .description("获取微信公众号文章正文（Markdown）")
     .option("-r, --referer <url>", "请求来源")
     .action(async (url, opts) => {
       try {

src/commands/mcp.ts

@@ -123,7 +123,7 @@ export function createMcpServer(): Server {
       },
       {
         name: "content",
-        description: "获取微信公众号文章正文",
+        description: "获取微信公众号文章正文（Markdown）",
         inputSchema: {
           type: "object" as const,
           properties: {

src/lib/__test__/fetcher.test.ts

@@ -102,4 +102,60 @@ describe("fetchArticleContent", () => {
       "获取文章内容失败: 未拿到有效的微信公众号文章链接",
     )
   })
+
+  it("正文按 Markdown 返回，保留标题、列表、代码块、图片、引用与表格", async () => {
+    const html = readFileSync(join(fixtureDir, "article-content.html"), "utf-8")
+    vi.stubGlobal("fetch", vi.fn().mockResolvedValue(mockRes("https://mp.weixin.qq.com/s/a", html)))
+
+    const md = await fetchArticleContent("https://mp.weixin.qq.com/s/a")
+
+    expect(md).toContain("### 小标题")
+    expect(md).toContain("第一段**加粗**文字。")
+    expect(md).toContain("-   条目一")
+    expect(md).toContain("> 引用一段话")
+    expect(md).toContain("```\nnpm install -g mpget\n```")
+    expect(md).toContain("| 列 A | 列 B |")
+    expect(md).toContain("末尾*斜体*结束。")
+  })
+
+  it("图片取 data-src，跳过 data: 占位图", async () => {
+    const html = readFileSync(join(fixtureDir, "article-content.html"), "utf-8")
+    vi.stubGlobal("fetch", vi.fn().mockResolvedValue(mockRes("https://mp.weixin.qq.com/s/b", html)))
+
+    const md = await fetchArticleContent("https://mp.weixin.qq.com/s/b")
+
+    expect(md).toContain("![示意图](https://mmbiz.qpic.cn/mmbiz_png/abc/0?wx_fmt=png)")
+    expect(md).not.toContain("data:image/gif")
+  })
+
+  it("图片地址与 alt 按 Markdown 语法转义", async () => {
+    const html = readFileSync(join(fixtureDir, "article-content.html"), "utf-8")
+    vi.stubGlobal("fetch", vi.fn().mockResolvedValue(mockRes("https://mp.weixin.qq.com/s/e", html)))
+
+    const md = await fetchArticleContent("https://mp.weixin.qq.com/s/e")
+
+    // 地址含空格与圆括号时 turndown 会加尖括号并转义，alt 里的方括号也会被转义
+    expect(md).toContain("![图\\[1\\]](<https://mmbiz.qpic.cn/a/b \\(1\\).png>)")
+  })
+
+  it("排除 script/style 与 js_content 之外的内容", async () => {
+    const html = readFileSync(join(fixtureDir, "article-content.html"), "utf-8")
+    vi.stubGlobal("fetch", vi.fn().mockResolvedValue(mockRes("https://mp.weixin.qq.com/s/c", html)))
+
+    const md = await fetchArticleContent("https://mp.weixin.qq.com/s/c")
+
+    expect(md).not.toContain("脚本内容不该出现在正文里")
+    expect(md).not.toContain("display: none")
+    expect(md).not.toContain("正文之外的噪声")
+  })
+
+  it("没有 js_content 时返回失败提示", async () => {
+    vi.stubGlobal(
+      "fetch",
+      vi.fn().mockResolvedValue(mockRes("https://mp.weixin.qq.com/s/d", "<div>没有正文</div>")),
+    )
+    expect(await fetchArticleContent("https://mp.weixin.qq.com/s/d")).toBe(
+      "获取文章内容失败: 正文为空",
+    )
+  })
 })