PDF批处理·全能转换：在 Agent 里做本地 PDF 批处理
PDF批处理·全能转换是长沙煋旺智能科技有限公司在 SkillHub 发布的企业认证 Skill（@org-l7548nxs/pdf-batch-master，v1.0.7）。它将 PDF 与 Office 互转、合并拆分、压缩、水印、加解密、表单、OCR 与质量检查等 20+ 能力封装为 Agent 可调用流程，标签 requires_api_key=false，强调纯本地与全功能免费。本文说明其定位、已核实功能、通过 skillhub CLI 或安装提示启用，以及 Cursor、Claude Code 等常见 skills 目录与对话式典型用法，并提醒安装路径与版式/OCR 复核等注意点。
Read More本地文件解析专家：用 LiteParse 在本地提取 PDF 与 Office 文档
SkillHub Skill「本地文件解析专家」基于开源 LiteParse，在本地解析 PDF、Word、PPT、Excel 与图片，支持纯文本、带坐标的 JSON、页面截图与目录批量处理，无需联网与文档云 API 费用。文章说明 Skill 定位、核心能力、skillhub install 安装方式、install.sh 与 pip 依赖，以及 lit parse、batch-parse、screenshot 等典型命令与 Python 调用，并列出 LibreOffice、OCR 与加密 PDF 等注意事项。
Read More处理发票：本地 OCR 提取金额与抬头
「处理发票」是 SkillHub 社区 Skill（@user_adcdbfa1/invoice-processor，v1.0.0），分类 office-efficiency，标注 requires_api_key=false。它在本地对发票做 OCR，提取金额、抬头等信息并支持多格式导出，数据不上传外部服务。安装需按 skillhub.cn 文档使用 skillhub install 并指定 Agent 的 skills 目录；启用后在对话发送「处理发票」即可。适合报销、对账等需本地处理财务影像的场景，具体导出格式与识别效果以 SKILL.md 与实际样本为准。
Read Moredsh-ocr-vision: A Local RapidOCR Text Extraction Tool
`dsh-ocr-vision` is a plugin for DeepSeek Harness designed to address the image understanding limitations of pure text models. It adds a local `ocr_image` tool that uses the RapidOCR engine to convert images into plain text, enabling the model to “understand” images. The plugin can be installed directly from GitHub or built locally, and requires Python and the RapidOCR dependencies to be installed on the host machine. Functionally, it supports reading images via `ctx.fs`, region cropping, and parameter configuration. Once installed, the model can directly invoke this tool to perform localized, privacy-preserving OCR processing without uploading images to any API.
Read Moredsh-docs: Local Document Parsing and OCR Plugin
dsh-docs is a localized document processing plugin designed specifically for DeepSeek Harness, enabling offline OCR recognition and text extraction without internet access. It supports reading multiple formats, including PDFs, Office documents, and images, and can convert content into Markdown or plain text. The plugin also provides full-text search and batch processing capabilities. Running on Windows, it emphasizes data privacy and security, making it suitable for offline environments or for handling sensitive documents.
Read More@zenk/vision: Enable DSH to directly process screenshots, keeping images on-device.
## Introduction There are two common approaches to showing an AI a screenshot: either manually converting the image content to text descriptions, or uploading the image to a cloud-based multimodal API. The former is time-consuming and often loses details such as layout and positioning; the latter raises concerns regarding privacy and cost. DSH's philosophy is "everything is a plugin," and visual capabilities can be integrated in the same manner. This article introduces @zenk/vision, which accomplishes exactly that: simply paste a screenshot directly into the chat interface, and the AI can view and analyze it without the image leaving your local machine. ## What is it @zenk/vision is a local visual plugin for DSH, maintained by kaaaaahn, open-sourced under the MIT license, and currently at version 0.3.1. It connects two local visual pipelines to DSH: macOS Visio
Read MoreLocalLens: Use macOS Vision OCR to locally read screen and image text for DeepSeek Harness
LocalLens is a community plugin for DeepSeek Harness (DSH) based on the macOS Vision framework, designed to perform OCR recognition locally. It extracts text from screen screenshots, clipboard images, or local files without requiring visual models, API keys, or network requests, returning the results to DSH for processing. The plugin supports macOS 13+ and Node.js 22+ environments, offering three input sources: `source=screen`, `clipboard`, and `file`, with default support for Simplified Chinese and English. Once installed, it can be invoked directly via natural language instructions—for example, to read clipboard content or parse text from local images—and supports subsequent data transformation and analysis. Please note that this plugin only processes bitmaps; PDFs must first...
Read Moredsh-tool-eyes: Add local vision tools for DeepSeek Harness
`dsh-tool-eyes` is a local visual plugin for DeepSeek Harness (DSH), designed to provide screen reading and image description capabilities for DSH agents in Windows environments. Licensed under the MIT license, this plugin provides two core tools: `screen` and `ocr`. The `screen` tool describes screen captures or specified images by invoking local OpenAI-compatible vision endpoints, supporting the use of local models to ensure data remains local and reduce privacy risks. The `ocr` tool utilizes the Windows built-in OCR engine to extract text from screens or images locally and offline, character by character, without requiring GPU or cloud support. Installation requires Windows 10/11, Node.js >
Read More