x4-ps

name	x4-ps
description	Capture clipboard image and let the agent view it directly with multimodal vision. Trigger with /x4:ps.

X4 Clipboard Image Vision

Use this skill for /x4:ps when the user wants the agent to see an image quickly.

Trigger

• /x4:ps
• User asks: "看图", "解析剪贴板图片", "让 agent 看到图片"

Purpose

Capture the current clipboard image into a file, then inspect it with multimodal vision (not OCR by default).

Workflow (MANDATORY)

1. Run:

bash skill-scripts/copy-img-clip.sh

2. If command exits 1, stop and report failure (clipboard has no image or capture failed).
3. If command exits 0, read the stdout path (full image path).
4. Open that local image path with vision tool and analyze visual content.
5. Return:
   • image path
   • concise visual summary
   • any text/structure recognized visually (without claiming exact OCR-level precision)

Policy

• Default path source must come from script stdout.
• Do not run OCR unless user explicitly requests OCR.
• Keep analysis grounded in visible content only; do not fabricate hidden details.

Output

• Image: <full-path>
• Result: <what is visible>
• Confidence/limits: <if small or unclear regions exist>