vision-support skillA
vision-support is agent-read markdown (skill) from penfick/skills: 为非多模态模型（如 deepseek-v4-pro、GLM-5.1、mimo-v2.5-pro 等纯文本模型）提供图片识别能力。当主模型无法识别图片、用户发送了截图/设计稿/UI 截图需要分析、或者用户说'看看这张图'、'分析这个截图'、'这张图片有什么问题'时，自动触发此技能。也适用于用户粘贴了图片但当前模型不支持图片输入的任何场景。支持同时识别多张图片，通过配置多个识图模型实现主备回退。使用指令 /skill:vision-support 或 /vision 也可手动触发。铁律：本技能配置的模型仅用于图片内容识别，绝不参与主逻辑推理。注意：如果当前模型本身是多模态模型（如 Claude Sonnet 4、GPT-4o、Gemini 等可以直接识图的模型），不要使用此技能，直接让主模型识别即可。.
Indexed from public GitHub and served as immutable, content-addressed versions. Install it pinned to an exact SHA-256 with the mdr CLI, and every file is verified against the hash recorded here before it reaches your agent. The deterministic audit below grades the latest version, and the same file always earns the same grade.
What the file says
# Vision Support — 非多模态模型的图片识别桥接 > **铁律：本技能配置的所有模型仅用于图片内容识别，绝不参与主逻辑推理。** > 这些模型不会代替主模型做任何决策、分析或编码，它们只负责"看"图片然后把看到的内容用文字描述出来。 ## 什么时候使用此技能 - **用户在对话中附带了图片**，但当前模型不支持图片理解 - **用户提到截图/图片/界面/设计**："看看这个截图"、"界面有问题"、"这个设计稿" - **用户描述了一个视觉问题但说不清楚**："网页显示不对"、"布局乱了" - **agent 在工作中遇到图片文件**（PNG/JPG/WebP 等） - **通过指令 `/vision` 或 `/skill:vision-support` 手动触发** ## 首次使用 — 一键初始化 ```bash node SKILL_DIR/scripts/vision.mjs init ``` 交互式引导，只需三步： 1. **选 Provider** — 从预置的主流平台列表中选择 2. **填密钥** — 输入 API Key（或环境变量名） 3. **选模型** — 自动从 API 拉取可用模型列表供选择（如拉取失败则显示推荐列表） 支持的平台覆盖国内外主流： | 分类 | 平台 | |------|------| | 国际 | OpenAI、Google Gemini、Anthropic Claude、DeepSeek、Groq、Mistral、xAI (Grok)、OpenRouter、Fireworks AI | | 国内 | 通义千问 (Qwen VL)、智谱 GLM (GLM-4V)、Moonshot (Kimi)、阶跃星辰 (Step)、MiniMax、SiliconFlow (硅基流动)、小米 MiMo | | 本地 | Ollama、LM Studio | | 自定义 | 任何 OpenAI 兼容的第三方平台（自填 baseUrl） | ## 添加备用模型 ```bash node SKILL_DIR/scripts/vision.mjs config add ``` 同样的交互式引导，添加的模型作为 fallback 回退。主模型失败后自动尝试。 ## 所有配置命令 ```bash # 交互式 node SKILL_DIR/scripts/vision.mjs init # 初始化主模型 node SKILL_DIR/scripts/vision.mjs config add # 添加 fallback …
Read the whole file at its exact version.
How to install
mdr add penfick/skills/vision-support@git:20260513.b41a63fmdr add penfick/skills/vision-support@sha256:a95cd0138b3779b3Pin to a label to follow the author's releases, or to a sha256 to freeze the exact bytes forever. Either way the resolved hash is written to mdr.lock, and mdr install reproduces it on any machine.
[](https://markdownregistry.com/a/art_yt2uho5kdhg2adhf)
1 badge views in 30 days
Versions
| version | committed | commit | size | audit | |
|---|---|---|---|---|---|
| git:20260513.b41a63f latest | 2026-05-13 | b41a63f | 4,900 B | A | view · diff |
| git:20260513.5b619de | 2026-05-13 | 5b619de | 4,715 B | A | view |
Audit of the latest version
- pass: Frontmatter block present
- pass: Frontmatter declares a name
- pass: Frontmatter declares a description
- pass: Size between 200 bytes and 200 KB (4900 bytes)
- pass: No zero-width or bidi control characters
- pass: No instruction hidden inside an HTML comment
- pass: No link to an exfiltration or paste host
- pass: No credential-shaped string
- pass: No instruction to send local credentials anywhere
- pass: No text hidden with inline styles
- pass: No prompt-injection phrasing
- pass: No curl or wget piped into a shell
- pass: No recursive delete of root, home or parent
- pass: No instruction to read or print local credentials
- pass: No base64 blob over 200 characters
- pass: No link to a raw IP address
- pass: No script tag
Source
penfick/skills · 12 stars · license MIT · pushed 2026-09-20 · branch master
API
GET https://markdownregistry.com/api/v1/artifacts/art_yt2uho5kdhg2adhf GET https://markdownregistry.com/api/v1/resolve?ref=penfick/skills/vision-support GET https://markdownregistry.com/api/v1/blob/a95cd0138b3779b3487106f51638526de9fa7ab13ee6eb409448b66e6eb89c39
Your agent does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.