agent-vision-toolkitA80

# 本仓库是上游工具箱；进 DSH 用它以子模块方式维护的独立包
dsh plugin --profile web add @anionex/dsh-vision-toolkit
# Headless profile 同理：dsh plugin --profile headless add @anionex/dsh-vision-toolkit

This is the upstream, cross-host repository for the vision toolkit; native DSH support landed on 2026-08-13 through the independently maintained Anionex/dsh-vision-toolkit package, which this repo tracks as a git submodule — which is why the install command here points at that package rather than at this repository. What lands in DSH is ten structured visual tools: intent-aware image Q&A, grounding, detection, tracing, cropping, pixel diff, long-screenshot OCR, foreground extraction, dominant-colour analysis and HTML screenshots. Beyond the tools it wires into DSH Credentials, a managed isolated runtime, previewable Artifacts, a Web settings panel, and Agent-scoped progressive tool exposure — so all ten tools are not dumped into the model's request schema up front.

• ▸Ten structured visual tools: intent-aware Q&A, grounding, detection, tracing, cropping, pixel diff, long-screenshot OCR, foreground extraction, dominant colour, HTML screenshots
• ▸DSH support ships as the separate @anionex/dsh-vision-toolkit package tracked here as a submodule; works in Web and Headless profiles
• ▸Integrates DSH Credentials, a managed isolated runtime, previewable Artifacts and Agent-scoped progressive tool exposure

通用视觉工具箱上游仓，DSH 集成由配套的 dsh-vision-toolkit 承担；本仓面向全 agent 生态

Drop it into your README — it links back to this page.

Badge (inline)Copy Markdown

[![dshfind](https://dshfind.com/api/badge/Anionex/agent-vision-toolkit)](https://dshfind.com/en/plugins/Anionex/agent-vision-toolkit?ref=badge)

Card (440×200)Copy Markdown

[![dshfind](https://dshfind.com/api/card/Anionex/agent-vision-toolkit)](https://dshfind.com/en/plugins/Anionex/agent-vision-toolkit?ref=badge)