Doc Processor

智能文档生成引擎 - 样式感知 + 三层模板 + 动态分类 支持 Word/Excel/PDF/CSV/TXT 读取、写入、转换、合并、提取、模板填充 v2.7: 性能监控 + 用户文档 + 错误增强 v2.6: Excel 格式保持 + PDF 生成 v2.5: 批量处理 + 模板缓存 v2.4: 日志系统 +...

Security Scan

The skill's files, dependencies, and runtime instructions are consistent with a local document-processing tool and do not request unrelated credentials or network access, but review of the install script and a runtime sanity check is recommended before installation.

Purpose & Capability

The name/description (Word/Excel/PDF processing, template filling, batch processing) matches the files and APIs present (doc_processor.py, excel_style.py, pdf_generator, scripts). The declared system binaries (pdftotext, pdfinfo) are appropriate for PDF reading. Required Python dependencies (python-docx, openpyxl, pandas) align with the stated features.

Instruction Scope

SKILL.md instructs only local operations (reading/writing/converting/merging files) and to run setup.sh / scripts under the skill workspace. The codebase refers to a workspace path (~/.openclaw/workspace/ and user-templates) and to optional environment variables for logging (DOC_PROCESSOR_LOG_LEVEL, DOC_PROCESSOR_LOG_FILE) that are not listed as required in the manifest — this is benign but worth noting. No obvious instructions to read unrelated system files or to transmit data externally appear in the provided snippets; earlier AI/networking modules were explicitly removed in recent versions.

Install Mechanism

There is no remote one-step binary download in the manifest; installation is via a local setup.sh that runs pip install -r requirements.txt and the metadata suggests using system package managers for poppler. pip installs from PyPI (or user-specified mirrors) are the only network activity implied. This is standard for Python projects but carries the usual supply-chain risk; the install script should be inspected before running.

Credentials

The skill declares no required credentials or config paths. It does use/observe optional env vars for logging (DOC_PROCESSOR_LOG_LEVEL, DOC_PROCESSOR_LOG_FILE) and the SKILL.md shows examples of providing a PIP_INDEX_URL when running setup.sh. These environment usages are reasonable for configuration; they are not secret credentials and do not appear to be used for exfiltration. Still, DOC_PROCESSOR_LOG_FILE allows writing logs to a path under user control — verify that the path is safe.

Persistence & Privilege

The skill is not flagged always:true and is user-invocable only. It stores user templates within the skill's workspace (~/.openclaw/workspace/user-templates/) which is consistent with its purpose. There is no indication it modifies other skills or global agent configuration.

Assessment

This package appears to be a locally focused document-processing tool and is internally coherent with its description. Before installing: (1) open and inspect setup.sh (it will run pip and any commands on your machine); (2) run check_deps.py to verify expected system tools (pdftotext/pdfinfo) and Python modules; (3) install inside a virtualenv to limit scope; (4) verify the logging env vars (DOC_PROCESSOR_LOG_FILE) won't write logs to sensitive locations; (5) if you want to be extra cautious, review any omitted files (setup.sh and remaining code not shown in the truncated dump) to confirm there are no unexpected network calls or shell invocations. If you need the old AI features, note they were removed in v2.7.11 and are expected to be provided by the OpenClaw host instead.

文档处理 Skill (Doc Processor)

统一的本地文档处理工具，支持多种格式的读取、写入、转换、合并和数据提取。

支持格式

格式	扩展名	读取	写入	转换	合并	提取
PDF	.pdf	✅	❌	⚠️	❌	✅
Word	.docx	✅	✅	✅	✅	✅
Excel	.xlsx	✅	✅	✅	✅	✅
CSV	.csv	✅	✅	✅	-	✅
文本	.txt	✅	✅	✅	-	✅
Markdown	.md	✅	✅	✅	-	✅

注意:

• ❌ 不支持 .doc 和 .xls 旧格式（需先用 LibreOffice 转换）
• ✅ PDF 处理使用系统工具 poppler-utils

v2.7.11 重要变更

AI 功能已移除

v2.7.11 起，AI 功能（AI 摘要、AI 分析）已移除，原因：

1. 职责单一: doc-processor 专注文档处理
2. 架构优化: AI 能力由 OpenClaw 主程序统一提供
3. 简化配置: 无需单独配置 LLM_BASE_URL 等环境变量
4. 提升安全: 移除网络请求，消除 ClawHub 安全标记

如需 AI 功能，请使用 OpenClaw 主程序的 LLM 能力：

# 1. 使用 doc-processor 读取文档
content = doc_processor.read('report.docx')

# 2. 将内容发送给 OpenClaw 的 LLM 进行摘要/分析
# (通过 OpenClaw 的消息系统或 API)

受影响的 API

以下方法已移除或修改：

方法	原功能	新行为
ContentAdapter.__init__(ai_service_type)	AI 服务配置	参数已移除
DocumentProcessor.__init__(ai_service_type)	AI 服务配置	参数已移除
create_ai_service()	创建 AI 服务	抛出 NotImplementedError
summarize_document()	AI 摘要	抛出 NotImplementedError

迁移指南

原代码 (v2.7.10):

processor = DocumentProcessor(ai_service_type='hybrid')
summary = processor.summarize_document('file.docx')

新代码 (v2.7.11):

# 1. 读取文档
processor = DocumentProcessor()
content = processor.read('file.docx')

# 2. 使用 OpenClaw 主程序进行 AI 处理
# (具体方式取决于 OpenClaw 的实现)

安装

一键安装

cd ~/.openclaw/workspace/skills/doc-processor
./setup.sh

安装脚本说明:

• ✅ 标准 pip 操作，安装 Python 依赖
• ✅ 依赖列表：python-docx, openpyxl, pandas, python-dotenv
• ✅ 从官方 PyPI 或可信镜像源安装
• ✅ 无恶意代码，完全开源

详情见：SECURITY.md

手动安装

1. 安装系统依赖

# Linux (Debian/Ubuntu)
sudo apt install poppler-utils

# Linux (RHEL/CentOS)
sudo yum install poppler-utils

# macOS
brew install poppler

2. 安装 Python 依赖

# 使用默认官方源
./setup.sh

# 使用清华镜像源（中国大陆推荐）
./setup.sh https://pypi.tuna.tsinghua.edu.cn/simple

# 或使用环境变量
export PIP_INDEX_URL=https://pypi.tuna.tsinghua.edu.cn/simple
./setup.sh

# 其他镜像源
./setup.sh https://pypi.mirrors.ustc.edu.cn/simple

3. 验证安装

python3 check_deps.py

使用示例

获取文档信息

./scripts/doc-info.sh report.pdf

读取文档

# 读取 PDF
./scripts/doc-read.sh report.pdf

# 读取 Word（包含表格）
./scripts/doc-read.sh meeting.docx

# 读取 Excel（所有 sheet）
./scripts/doc-read.sh data.xlsx

写入文档

# 写入 Word
echo '{"title": "报告", "paragraphs": ["第一段", "第二段"]}' | \
  ./scripts/doc-write.sh -o output.docx

# 写入 Excel
echo '[["姓名", "年龄"], ["张三", 25], ["李四", 30]]' | \
  ./scripts/doc-write.sh -o data.xlsx

格式转换

# Word 转 CSV
./scripts/doc-convert.sh table.docx table.csv

# Excel 转 Word
./scripts/doc-convert.sh data.xlsx report.docx

# PDF 转文本
./scripts/doc-convert.sh report.pdf report.txt

合并文档

# 合并 Word 文档
./scripts/doc-merge.sh part1.docx part2.docx -o merged.docx

# 合并 Excel（每个文件→独立 sheet）
./scripts/doc-merge.sh q1.xlsx q2.xlsx -o yearly.xlsx

提取数据

# 提取 Word 中的表格
python3 doc_processor.py --action extract \
  --input report.docx --options '{"include_tables": true}'

# 提取 Excel 指定 sheet
python3 doc_processor.py --action extract \
  --input data.xlsx --options '{"sheet_names": ["Sheet1"]}'

Python API

from doc_processor import DocumentProcessor

processor = DocumentProcessor(workspace="/path/to/workspace")

# 获取文档信息
info = processor.get_info("report.docx")
print(info.to_dict())

# 读取内容
content = processor.read("report.docx")
print(content.data['paragraphs'])

# 写入文档
processor.write("output.docx", {
    "title": "新文档",
    "paragraphs": ["内容 1", "内容 2"]
})

# 转换格式
processor.convert("input.docx", "output.csv")

# 合并文档
processor.merge(["a.docx", "b.docx"], "merged.docx")

# 提取数据
data = processor.extract("data.xlsx")

依赖说明

依赖	类型	用途	自动安装
poppler-utils	系统	PDF 读取	✅ (Linux/macOS)
python3	系统	运行环境	❌ (需预装)
python-docx	Python	Word 处理	✅ (setup.sh)
openpyxl	Python	Excel 读写	✅ (setup.sh)
pandas	Python	数据处理	✅ (setup.sh)

更新日志

v1.0.4 (2026-03-25)

Bug 修复:

• ✅ 修复 Excel 写入失败问题 - 支持 {'sheet_name': '...', 'data': [...]} 格式
• ✅ 修复 Word 合并失败问题 - 正确处理相对路径，避免 "Package not found" 错误

v1.0.3 (2026-03-25)

Bug 修复:

• ✅ 修复 CSV→Excel 转换问题
• ✅ 修复 DocInfo 属性问题

常见问题

Q: 为什么不支持 .doc 和 .xls？

A: 这些是旧的二进制格式，需要 LibreOffice 转换。建议先用 LibreOffice 转换为 .docx/.xlsx：

libreoffice --headless --convert-to docx old.doc
libreoffice --headless --convert-to xlsx old.xls

Q: PDF 写入支持吗？

A: 不支持。PDF 是只读格式。如需生成 PDF，可先创建 Word 文档再手动转换。

Q: 大文件处理慢怎么办？

A: 建议使用 --options 限制读取范围：

# 只读取 PDF 前 5 页
python3 doc_processor.py --action read \
  --input large.pdf --options '{"pages": "1-5"}'

许可证

MIT License