海运托书智能提取 Shipping Booking Extractor

海运托书（Booking Note / Shipping Instruction）智能提取工具。从 PDF、图片、扫描件、Word、Excel、RTF 等格式的托书文件中提取结构化信息，输出标准 JSON。当用户明确提到"托书"、"booking"、"shipping instruction"、"帮我提取"、"解...

Security Scan

The skill's code, instructions, and requested credentials line up with its stated purpose (extracting structured data from shipping booking documents) and do not request unrelated privileges, though it will send uploaded documents to whichever third‑party LLM/vision API you configure.

Purpose & Capability

The skill's name/description match the included script and docs: it extracts text/images from PDFs, Word, Excel, RTF and calls an LLM/vision provider to produce JSON. The primary credential (Anthropic) and optional OpenAI-compatible keys are appropriate for that purpose. Minor metadata inconsistency: top-level 'Required env vars: none' vs SKILL.md metadata which declares it requires either ANTHROPIC_API_KEY or OPENAI_API_KEY.

Instruction Scope

Runtime instructions are narrowly scoped: run the provided extract.py on an uploaded file. The script reads/convert the given file(s), may call local helpers (textutil / libreoffice) for .doc/.rtf, encodes images as base64, and sends content to configured model endpoints. It does not instruct reading unrelated system files or credentials beyond the declared API keys. It will transmit the document contents as part of API requests (as described in the SKILL.md).

Install Mechanism

No opaque network install steps are present. Dependencies are standard Python packages listed in requirements.txt (anthropic, openai, pdfplumber, pymupdf, python-docx, openpyxl, xlrd, striprtf). The recommended pip install is proportionate to the document-processing goal; there are no downloads from arbitrary URLs or archive extraction operations in the manifest.

Credentials

Requested credentials align with functionality: ANTHROPIC_API_KEY is primary, and OPENAI_API_KEY is offered as an alternative to call OpenAI-compatible endpoints. The SKILL.md also references OPENAI_BASE_URL and optional SHIPPING_MODEL/SHIPPING_VISION_MODEL env vars for configuration — these are reasonable but should be noted. No unrelated secrets or system credentials are requested. There is a small inconsistency between registry 'Required env vars: none' and SKILL.md's declared anyEnv (ANTHROPIC_API_KEY or OPENAI_API_KEY).

Persistence & Privilege

The skill is not forced-always, and does not request elevated or persistent platform privileges. It runs as an on-demand script, may invoke local binaries (textutil/libreoffice) for format conversion, and writes temporary files during conversion — behavior consistent with its purpose.

Assessment

This skill appears to do what it says: it converts uploaded booking documents to text/images and sends them to the configured AI model(s) to extract JSON. Before installing, consider: - Data exposure: uploaded documents (text and embedded images) are included in API requests to third‑party model endpoints (Anthropic/OpenAI/etc.). If files contain sensitive commercial or personal data, restrict use or choose an on-prem / privacy-preserving model. - API key placement: the skill expects ANTHROPIC_API_KEY or OPENAI_API_KEY (and optionally OPENAI_BASE_URL or SHIPPING_MODEL/SHIPPING_VISION_MODEL). Verify where OpenClaw stores these keys and limit their scope/permissions where possible. - Local binaries: the script may call textutil (macOS) or libreoffice (headless) for .doc/.rtf conversion; ensure these tools are trusted and available if you need those formats. - Test with non-sensitive files first to confirm behavior. Also note the minor manifest inconsistency: the top-level metadata says no required env vars while the SKILL.md and script expect an LLM API key; this is likely harmless but you should ensure you provide the intended key before use.

Shipping Booking Extractor

从各种格式的海运托书文件中提取结构化数据，输出标准 JSON。

---

🚀 快速开始（首次安装）

第一步：安装 skill

clawhub install shipping-booking

第二步：安装依赖（只需一次）

pip install -r ~/.openclaw/skills/shipping-booking/requirements.txt

第二步：配置 AI 模型（任选一种，无需全部配置）

方式	说明
✅ 推荐	在 OpenClaw 中已配置 AI 账号，自动读取，无需任何操作
Anthropic Claude	export ANTHROPIC_API_KEY=your_key
DeepSeek	export OPENAI_API_KEY=your_key export OPENAI_BASE_URL=https://api.deepseek.com/v1
通义千问	export OPENAI_API_KEY=your_key export OPENAI_BASE_URL=https://dashscope.aliyuncs.com/compatible-mode/v1
Kimi	export OPENAI_API_KEY=your_key export OPENAI_BASE_URL=https://api.moonshot.cn/v1
智谱 GLM	export OPENAI_API_KEY=your_key export OPENAI_BASE_URL=https://open.bigmodel.cn/api/paas/v4
自定义文本模型	export SHIPPING_MODEL=your_model_name
自定义视觉模型	export SHIPPING_VISION_MODEL=your_vision_model（用于图片/PDF）

💡 只要在 OpenClaw 里配置过任意一个 AI 账号，装完即用，不需要手动设置环境变量。

第三步：直接使用，把托书文件发给我即可！

---

支持格式

格式	支持情况
PDF	✅ 全平台
图片（JPG/PNG/TIFF/BMP/WebP）	✅ 全平台
Excel（.xlsx / .xls）	✅ 全平台
Word（.docx）	✅ 全平台
RTF（.rtf）	✅ 全平台
Word（.doc）	✅ macOS 自动处理；其他平台需安装 LibreOffice 或另存为 .docx

---

触发条件

必须同时满足：

1. 用户上传了文件
2. 用户提到以下任意关键词：
   • 中文：托书、提单、订舱、帮我提取、解析、识别
   • 英文：booking、shipping instruction、SI、extract

---

执行流程

Step 1：确认文件类型，运行提取脚本

python3 ~/.openclaw/skills/shipping-booking/scripts/extract.py <文件路径>

Step 2：验证并输出结果

• 提取成功 → 将 JSON 以代码块形式呈现，不要暴露终端命令
• low_confidence_fields 不为空 → 在 JSON 下方附加提醒：

  ⚠️ 以下字段置信度较低，建议人工核查：字段1、字段2

• 不像托书文件 → 提示：

  ❌ 这不像是托书文件，请确认上传的文件是否正确

---

🔒 数据安全说明

• 托书文件内容（文本/图片）会发送给你配置的 AI 模型 API 进行分析
• 支持的 API 端点：Anthropic Claude、OpenAI、DeepSeek、通义千问、Kimi、智谱 GLM
• 文件本身不会上传到任何第三方存储，仅作为 API 请求内容临时传输
• API Key 从本地 OpenClaw 配置读取，不会泄露到其他地方

字段定义

详见 references/schema.md