vision-subagent SKILL.md skillA
vision-subagent from orinvoss/dsh-math-team is an agent skill: a SKILL.md folder an AI agent loads when a task calls for it.
mdr add orinvoss/dsh-math-team/vision-subagent@git:20260910.1313758Pins git:20260910.1313758 in mdr.lock. Pin the fingerprint instead.
- Audit grade
- A, 17 of 17 checks passed
- Kind
- SKILL.md skill
- Repository
- orinvoss/dsh-math-team
- Path
skills/vision-subagent/SKILL.md- Versions recorded
- 8
- Latest version
- git:20260910.1313758, committed Sep 10, 2026
- First version recorded
- Aug 14, 2026
- Size
- 6,535 bytes
- Repository stars
- 7
- License
- MIT
视觉子代理。当主模型不支持图像输入时，用 workflow 派生一个指定当前环境可用视觉模型的子代理，通过 read_image 查看或核验图片内容并结构化返回结论。适用于"需要看图、但主模型读不了图"的任何会话。.
Collected from public GitHub. Every version is saved exactly as it was and never changes. Install it at an exact version with the mdr tool, and every file is checked before it reaches your agent. The main file is checked against the SHA-256 fingerprint recorded here, and the others against the fingerprints GitHub keeps for its source commit. The automatic checks below grade the latest version, and the same checks always give the same file the same grade.
What the file says
# 视觉子代理（Vision Sub-agent）
当**主模型不支持图像输入**（`read_image` 会拒读，如纯文本模型）时，用 **`workflow` 派生一个指定视觉模型的子代理**来"看"图——它可以用 `read_image` 读取图片，再根据你的要求描述内容或审查质量并返回结论。
本 skill 是**全局可用**的看图方案，任何预设/会话/项目都可加载，不限定业务场景。
## 触发条件（先判断，再决定用不用本 skill）
**先试一下**：直接调 `read_image`，或查 `llm.resolveModelInfo(provider, model)` 的 `inputModalities`。
- ✅ **主模型支持图像**（`inputModalities` 含 `image`）→ **直接 `read_image` 读图即可，不必派子代理**（更快、少一层）。注意能力靠配置声明——若报 `does not declare image input`，先按本文末"图像能力靠配置声明"排查，可能只是配置漏声明。
- ❌ **主模型不支持图像**（`read_image` 拒读）→ 用本 skill 派生视觉子代理来看图。
- 🔀 **仍值得用子代理的两种情形**：① 想要**独立/隔离的看图结论**（不让主模型既产出又自查）；② 想用**更便宜的视觉模型**替代主模型读图。
## 原理
DeepSeek Harness 的 `workflow` 工具允许在 `agent()` 上独立指定 `provider` 与 `model`。只要选一个**当前部署里真正支持图像输入的模型**（其 `inputModalities` 含 `image`），派生的子代理就能用 `read_image` 读取并处理图片。
不同部署注册的视觉模型名可能不同，**不要硬编码模型名**，应先探测。
## 如何找可用的视觉模型
用 `llm` 服务遍历各 provider 的模型，过滤出 `inputModalities` 含 `image` 的：
```js
// 思路：遍历 provider -> resolveModelInfo -> 过滤 inputModalities 含 'image'，得到 { provider, model }
const vision = /* inputModalities 含 'image' 的 {provider, model} */
await agent(prompt, { provider: vision.provider, model: vision.model })
```
…Read the whole file at its exact version.
How to install
mdr add orinvoss/dsh-math-team/vision-subagent@sha256:c0e8304a61377a4aThe line at the top of this page pins the label git:20260910.1313758, which is what most people want. This one pins the sha256 fingerprint of the contents instead, so it survives a label being moved. Either way the exact version is written to mdr.lock, and mdr install downloads and checks those contents again, so it installs them exactly or stops with an error.
[](https://markdownregistry.com/a/art_3663sx37da5by5th)
4 badge image fetches in 30 days. What the badge shows.
Versions
The registry has recorded 8 versions of this file, between Aug 14, 2026 and Sep 10, 2026, about one every 4 days.
| version | committed | commit | size | audit | |
|---|---|---|---|---|---|
| git:20260910.1313758 latest | 2026-09-10 | 1313758 | 6,535 B | A | view · diff |
| git:20260910.5c72975 | 2026-09-10 | 5c72975 | 5,964 B | A | view · diff |
| git:20260910.fa29fae | 2026-09-10 | fa29fae | 5,569 B | A | view · diff |
| git:20260822.c37c6cc | 2026-08-22 | c37c6cc | 5,557 B | A | view · diff |
| git:20260822.d100f0f | 2026-08-22 | d100f0f | 4,715 B | A | view · diff |
| git:20260822.6d41464 | 2026-08-22 | 6d41464 | 4,824 B | A | view · diff |
| git:20260814.c898c14 | 2026-08-14 | c898c14 | 3,367 B | A | view · diff |
| git:20260814.9d677ae | 2026-08-14 | 9d677ae | 3,589 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 (6535 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
orinvoss/dsh-math-team · 7 stars · license MIT · pushed 2026-09-29 · branch main
API
GET https://markdownregistry.com/api/v1/artifacts/art_3663sx37da5by5th GET https://markdownregistry.com/api/v1/resolve?ref=orinvoss/dsh-math-team/vision-subagent GET https://markdownregistry.com/api/v1/blob/c0e8304a61377a4ab6b99e3977788da1d00a32a395b3b19f699d6f1e3f1dd4e7
Your AI assistant 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.
Guides: the SKILL.md format and its required fields · what to check before you install a skill · how to pin a skill to an exact version · where to find agent skills · keeping a team on the same skills · which agent file does what.