Agent Skills › espressif/esp-claw › cap_llm_inspect_image

cap_llm_inspect_image

GitHub

用于分析设备上本地图片的视觉内容。当用户询问图片、照片或截图中的物体、文字或场景时触发，需传入绝对路径和具体提示词进行图像识别与描述。

components/claw_capabilities/cap_llm_inspect/skills/cap_llm_inspect_image/SKILL.md espressif/esp-claw

触发场景

用户询问图片中有什么需要分析本地照片或截图的视觉内容

安装

npx skills add espressif/esp-claw --skill cap_llm_inspect_image -g -y

更多选项

非标准路径

npx skills add https://github.com/espressif/esp-claw/tree/master/components/claw_capabilities/cap_llm_inspect/skills/cap_llm_inspect_image -g -y

不安装直接使用

npx skills use espressif/esp-claw@cap_llm_inspect_image

指定 Agent (Claude Code)

npx skills add espressif/esp-claw --skill cap_llm_inspect_image -a claude-code -g -y

安装 repo 全部 skill

npx skills add espressif/esp-claw --all -g -y

预览 repo 内 skill

npx skills add espressif/esp-claw --list

SKILL.md

Frontmatter

{
    "name": "cap_llm_inspect_image",
    "metadata": {
        "cap_groups": [
            "cap_llm_inspect"
        ],
        "manage_mode": "readonly"
    },
    "description": "How to inspect a local image with inspect_image."
}

Image Inspection

Use this skill when the user wants the device to inspect a local image and describe what is visible.

When to use

The user asks what is in an image, photo, screenshot, or camera frame.
The image already exists on the device filesystem.
The task needs visual analysis rather than plain file reading.

Available capability

inspect_image: analyze one local image from an absolute path using a prompt that says what to inspect.

Calling rules

Call inspect_image directly.
Always pass an absolute local file path in path.
Always pass a clear prompt that tells the model what to look for.
Confirm or discover the image path first if it is not already known.
Do not pass remote URLs or non-image files.

Path guidance

Prefer real local paths already stored on the device.
If the exact path is unknown, inspect storage first with file capabilities such as list_dir.
Common roots in this demo include <storage_root>/inbox, <storage_root>, or other application-managed storage paths.

Example

{
  "path": "<storage_root>/inbox/photo.jpg",
  "prompt": "Describe the main objects in this image and mention any visible text."
}

Notes

Keep the prompt specific. For example: identify objects, read visible text, describe a scene, or check whether a target item appears.
If the image is blurry or uncertain, report that uncertainty instead of over-claiming.