feishu-doc-extended

飞书文档扩展工具，提供图片下载和 OCR 识别功能。需要配合内置 feishu 插件使用。

Security Scan

The skill's declared functionality (download image URLs from Feishu and run OCR) is plausible and the tesseract install is appropriate, but it asks you to modify an internal OpenClaw/feishu plugin in-place (and the SKILL.md and included files mismatch), which is an elevated and undocumented change and warrants caution.

Purpose & Capability

The skill's purpose (fetch image URL from Feishu + OCR) matches required binary tesseract. However, the runtime instructions require editing OpenClaw's built-in feishu extension sources under /usr/local/lib/node_modules/openclaw/extensions/feishu — a system-level change not declared in 'required config paths'. Also the packaged files (src/getImage.ts) don't exactly match the filenames referenced in the SKILL.md (docx.ts), creating ambiguity about what to change.

Instruction Scope

SKILL.md tells the operator/agent to modify other extension source files (schema and docx handlers), add a new action, and restart the gateway. That goes beyond a normal skill's runtime instructions (which usually call APIs or run binaries) and requires write access to node_modules and restarting a system component. The instructions are specific about paths, so they will cause persistent code changes to platform files if followed.

Install Mechanism

Install steps are limited to brew installing tesseract and language data (tesseract-lang), which is a standard, low-risk package install on macOS. There is no arbitrary URL download or extracted archive in the install spec.

Credentials

The skill itself does not request environment variables or secrets. However, the added getImage implementation calls the existing feishu plugin's token manager (getTenantAccessToken) and client.httpInstance, meaning it relies on the feishu extension's credentials at runtime. That is expected for the stated purpose but implies access to tenant tokens via the modified plugin.

Persistence & Privilege

The skill requires persistent, in-place modification of the built-in feishu extension source code and a gateway restart. Those changes are persistent across runs and affect platform code beyond this skill, but the manifest does not declare any required config paths or elevated privileges. This persistent modification increases blast radius if the change is incorrect or malicious.

What to consider before installing

This skill does what it says (fetch Feishu image URLs and use tesseract for OCR), but it requires you to edit OpenClaw's built-in feishu extension code and restart the gateway — a sensitive, persistent change. Before installing: (1) review the exact code changes (the repo includes getImage.ts and doc-schema.ts — compare these to the actual files on your system); (2) prefer contributing the change upstream or implement a safer plugin hook rather than editing node_modules directly; (3) back up the original files and test in a non-production environment; (4) verify the getImage code only calls api endpoints under open.feishu.cn and does not exfiltrate tokens; (5) be cautious because the SKILL.md filenames and included files do not exactly match, which increases risk of accidental mis-modification. If you are not comfortable making these persistent platform edits, do not install or run the modification steps.

飞书文档扩展工具，提供图片下载和 OCR 识别功能。

功能

功能	说明
get_image	获取飞书文档中图片的下载 URL
image_ocr	下载图片并进行 OCR 文字识别（需要 tesseract）

依赖

• tesseract + 中文语言包（用于 OCR）
• 飞书开放平台应用权限
• OpenClaw 内置 feishu 插件

安装

# 安装 tesseract
brew install tesseract

# 安装中文语言包
brew install tesseract-lang

修改内置插件

本技能需要修改 OpenClaw 内置的 feishu 插件：

1. 修改 doc-schema.ts

文件路径: /usr/local/lib/node_modules/openclaw/extensions/feishu/src/doc-schema.ts

在 FeishuDocSchema 的 Union 类型末尾添加：

// Image download
Type.Object({
  action: Type.Literal("get_image"),
  image_token: Type.String({ description: "Image token (from block image.token)" }),
}),

2. 修改 docx.ts

文件路径: /usr/local/lib/node_modules/openclaw/extensions/feishu/src/docx.ts

1. 在文件末尾（uploadFileBlock 函数后）添加：

async function getImage(client: Lark.Client, imageToken: string) {
  const domain = client.domain ?? "https://open.feishu.cn";
  const token = await client.tokenManager.getTenantAccessToken();

  const res = await client.httpInstance.get<{ code?: number; data?: { image_url?: string } }>(
    `${domain}/open-apis/image/v4/get`,
    {
      params: { image_token: imageToken },
      headers: { Authorization: `Bearer ${token}` },
    },
  );

  if (res.data?.code !== 0 && res.data?.code !== undefined) {
    throw new Error(`Failed to get image: ${res.data}`);
  }

  return {
    image_url: res.data?.data?.image_url,
    image_token: imageToken,
  };
}

2. 在 switch 语句中添加 case：

case "get_image":
  return json(await getImage(client, p.image_token));

3. 重启 Gateway

openclaw gateway restart

使用方法

1. 获取文档中的图片 token

使用 feishu_doc 工具的 list_blocks 获取文档中的图片 block：

{
  "action": "list_blocks",
  "doc_token": "文档Token"
}

从返回结果中获取图片的 token（在 block.image.token 中）。

2. 获取图片下载 URL

{
  "action": "get_image",
  "image_token": "图片Token"
}

返回：

{
  "image_url": "https://xxx...",
  "image_token": "图片Token"
}

3. OCR 识别

获取图片 URL 后，可以用浏览器打开并截图，然后用 tesseract 识别：

tesseract /path/to/screenshot.jpg - -l chi_sim

工作流程

1. feishu_doc list_blocks → 获取图片 block 和 token
2. feishu_doc get_image → 获取图片下载 URL
3. 浏览器访问 URL → 截图
4. tesseract OCR → 识别文字

注意事项

• get_image 返回的 URL 是飞书临时 URL，有时效性
• 如果 URL 过期，需要重新调用 get_image
• OCR 识别效果取决于图片清晰度

更新日志

• 2026-03-12: 初始版本，添加 get_image 功能