Vision tools · DeepSeek Harness plugins — Page 51

DeepSeek Harness plugin: drive your local Adobe Photoshop from the agent — batch subject cutouts (Select Subject / Remove Background) to transparent PNGs, plus a raw ExtendScript escape hatch. · 让 AI 直接驱动本机 Photoshop 批量抠图。

DeepSeek Harness Web 插件：把 plantuml / puml 代码块渲染为图片，支持点开看大图、复制 JPG、复制 PlantUML 源码

Configure model input capabilities and reasoning efforts for DeepSeek Harness

DSH 截图提问插件：输入框相机按钮或 Ctrl+Shift+A → 自动最小化窗口 → 原生全屏框选浮层（确认/取消按钮）→ 选区直接作为图片附件进入当前会话输入框，全程不落盘。Screenshot-to-ask for DeepSeek Harness: the window hides, a native region overlay covers every monitor, and the crop lands in the composer as an image attachment.

Render local image files inline in the DSH conversation via a global show_image tool.

DeepSeek Harness 表情包插件：AI 在聊天时根据情境挑选并发送表情包图片（蓝发鱼娘/蓝色大肥鱼系列 109 张），聊天内直接内嵌显示。

Plugin-author upgrade skill for DeepSeek Harness: the closed 0.1.5-rc.2 -> 0.1.6-alpha.2 corridor card plus a zero-dependency seam scanner over one five-seam catalog - E1 (agent/created listener throwing blocks creation), E2 (async apply race: registration after the first await), E3 (removed slot keys settings.plugin.item and SessionListState.current), E4 (removed client APIs sessions.open/openSubagent/clear) and E5 (removed model literals deepseek-v4-flash*/deepseek-v4-vision-exp) - packaged as a bundle skill and an npx CLI.

Use image-generation HTTP APIs as native image models in DeepSeek Harness

给 DSH agent 装一双眼睛：调用在线 VLM（多供应商、OpenAI 兼容）分析本地图片、URL 与会话上传的附件；勾选的模型特色能力会实时注入系统提示

Vision capability plugin for DeepSeek Harness — enables pure-text LLMs to analyze images via GLM-4V-Flash

Vision assist for DeepSeek Harness: lets the main model see images through an OpenAI-compatible vision API

DeepSeek Harness native plugin: give text-only LLMs (e.g. DeepSeek) vision via free Gemini / GLM vision APIs — describe images, OCR, VQA, screenshots, charts.

Auxiliary vision for DeepSeek Harness: analyze images via an OpenAI-compatible vision endpoint (works with any main model)

Vision bridge for the DeepSeek Harness web GUI: upload/paste images into a chat, describe them through an OpenAI-compatible vision API, and feed English text descriptions to text-only models.

Vision bridge: read_image is transparently served by a configurable vision model (default zai/glm-4.6v) when the session model is text-only; the description returns as text and the session keeps its selected model. Ships a live on/off card in Settings - Plugins

DeepSeek Harness Web wallpaper plugin: upload and crop a local image as the web workspace desktop background, with a dedicated Settings section.

Wallpaper switcher for DeepSeek Harness — sidebar picker with local image upload, dim slider, and a one-click next-wallpaper button. · DSH 壁纸切换：页面上直接添加本地图片壁纸、遮罩浓度滑块（0–100%）、一键切换下一张。

DSH WebUI background plugin: pick a local image as the webui background, overlaid with a theme-matched translucent mask so it never disturbs the UI. Persistence is browser-local (localStorage).

DSH web plugin: serve workspace image files over a constrained /dsh-image route so agents can embed images inline in conversation.

Bridges the four z.ai GLM Coding Plan MCP servers into DeepSeek Harness via the in-box @deepseek-ai/dsh-mcp-client: vision (GLM-4.6V stdio via npx @z_ai/mcp-server), web reader, web search prime, and zread repo MCP (streamable-http at open.bigmodel.cn)

DSH 增强插件：视觉/记忆设置面板 + 插件市场（+ 后续挂机游戏）

DeepSeek Harness web plugin — upload files and folders straight into the conversation (no workspace copy needed): images attach as real image blocks, text content is written into the composer, folders land under ~/.dsh/uploads by directory structure

中文 PPT 生成技能：结构化大纲 → 视觉设计 → HTML/PPTX 出稿，附排版自查清单。原创技能（参考 GordenPPTSkill / ppt-master 思路）。

Eyes for your DeepSeek Harness agent — it opens, sees, and fixes what it builds. Headless-browser preview & self-verification tools (screenshot, console, DOM read, interact) as a dsh plugin.