← 返回 Skills 市场
mar-computer-vision-expert
作者
MarjorieBroad
· GitHub ↗
· v1.0.0
· MIT-0
73
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install mar-computer-vision-expert
功能描述
SOTA Computer Vision Expert (2026). Specialized in YOLO26, Segment Anything 3 (SAM 3), Vision Language Models, and real-time spatial analysis.
安全使用建议
This skill will read local image files and send their base64 contents to a third-party API (https://api.heybossai.com) and requires an API key (SKILLBOSS_API_KEY) according to SKILL.md — but the registry metadata does not declare that credential. Before installing, consider: 1) Do you trust the heybossai.com endpoint and its privacy policy for any images you might send? 2) Prefer to supply a limited-scope/test API key (not production secrets) and try only non-sensitive images first. 3) Request the publisher to update registry metadata to explicitly list SKILLBOSS_API_KEY and provide a homepage or source repo so you can audit the implementation. 4) If you must analyze sensitive imagery, run models locally or use a vetted enterprise VLM with clear data-handling guarantees. If the publisher cannot provide a trustworthy source repo or explain the missing credential declaration, treat the package with extra caution.
能力标签
能力评估
Purpose & Capability
The skill's stated purpose (YOLO26, SAM 3, VLM integration) matches the instructions which call a remote VLM hub, but the registry metadata claims no required environment variables or credentials while the SKILL.md explicitly uses SKILLBOSS_API_KEY to call https://api.heybossai.com. That omission is an incoherence: a VLM-proxy skill legitimately needs an API key, yet the metadata does not declare it.
Instruction Scope
SKILL.md contains a Python example that reads a local file (image.jpg), base64-encodes it, and POSTs it to api.heybossai.com/pilot. Reading local images and sending them to an external VLM service is expected for visual QA, but it is effectively exfiltration of image contents to a third party. The instructions do not limit what files may be read or warn about privacy/sensitivity, and they hard-code an example file read pattern (open('image.jpg')).
Install Mechanism
Instruction-only skill with no install spec or code files — nothing is written to disk or pulled at install time, which minimizes install-time risk.
Credentials
The SKILL.md requires SKILLBOSS_API_KEY for the SkillBoss API Hub; requiring a single API key for an external VLM service is proportionate. However, the registry metadata declared 'Required env vars: none' and 'Primary credential: none', creating a mismatch. The missing declaration reduces transparency about what secrets are needed and where they will be used.
Persistence & Privilege
No 'always' flag, no install-time scripts, and no declared config paths. The skill does not request persistent system presence or broader agent privileges.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install mar-computer-vision-expert - 安装完成后，直接呼叫该 Skill 的名称或使用
/mar-computer-vision-expert触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v1.0.0
computer-vision-expert v1.0.0 initial release
- Introduces expert guidance on SOTA (2026) computer vision: YOLO26, SAM 3, and leading Vision Language Models.
- Details advanced capabilities: real-time detection (NMS-free), promptable segmentation, 3D scene reconstruction, and VLM-powered reasoning.
- Provides best practices for edge deployment, API integration (SkillBoss API Hub), and vision pipeline design.
- Lists anti-patterns, potential pitfalls, and modern solutions for spatial AI challenges.
- Highlights compatibility and deployment considerations for real-world industrial and IoT use cases.
元数据
常见问题
mar-computer-vision-expert 是什么？
SOTA Computer Vision Expert (2026). Specialized in YOLO26, Segment Anything 3 (SAM 3), Vision Language Models, and real-time spatial analysis. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 73 次。
如何安装 mar-computer-vision-expert？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install mar-computer-vision-expert」即可一键安装，无需额外配置。
mar-computer-vision-expert 是免费的吗？
是的，mar-computer-vision-expert 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
mar-computer-vision-expert 支持哪些平台？
mar-computer-vision-expert 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 mar-computer-vision-expert？
由 MarjorieBroad（@marjoriebroad）开发并维护，当前版本 v1.0.0。
推荐 Skills