AgentSkillsCN

gemini-image-analysis

通过generateContent REST API，利用Gemini Flash（视觉模型）分析单张或多张图像。借助Bun + TypeScript对图像进行Base64编码，并输出模型预测结果。可用于生成图片说明、类似OCR的文字提取、UI/截图分析，以及多张图像的对比分析。

SKILL.md
--- frontmatter
name: gemini-image-analysis
description: Analyze one or multiple images with Gemini Flash (vision) via the generateContent REST API. Uses Bun + TypeScript to base64-encode images and print model output. Use for captioning, OCR-like extraction, UI/screenshot analysis, and multi-image comparison.

Gemini Image Analysis (Bun + TypeScript)

When to use

Use this Skill when you need to analyze one or multiple images (screenshots, photos, diagrams) with Gemini Flash via the REST API.

Quick start

bash
GEMINI_API_KEY="YOUR_KEY" \
bun .claude/skills/gemini-image-analysis/scripts/gemini-image-analyze.ts \
  --prompt "Caption this image." \
  /path/to/image.jpg

macOS Keychain (optional)

If you store GEMINI_API_KEY in the macOS Keychain:

bash
GEMINI_API_KEY="$(security find-generic-password -a "$(whoami)" -s "GEMINI_API_KEY" -w)" \
bun .claude/skills/gemini-image-analysis/scripts/gemini-image-analyze.ts \
  --prompt "Caption this image." \
  /path/to/image.jpg

Multiple images

Send multiple images in one request (useful for comparison or “before/after”):

bash
GEMINI_API_KEY="YOUR_KEY" \
bun .claude/skills/gemini-image-analysis/scripts/gemini-image-analyze.ts \
  --prompt "Compare these images and describe the differences." \
  /path/to/image1.jpg \
  /path/to/image2.png

Options

  • •--prompt <text>: Text instruction for the model (default: Caption this image.)
  • •--model <model>: Model name (default: gemini-3-flash-preview)
  • •--json: Print the full JSON response instead of extracted text

Notes

  • •The script reads GEMINI_API_KEY from the environment and never stores it.
  • •Supported image types: .jpg, .jpeg, .png, .webp, .gif.