AgentStack
SKILL verified Apache-2.0 Self-run

Hello Multimodal

skill-hellowind777-hello-multimodal-hello-multimodal · by hellowind777

此技能应在用户消息包含图片/截图/照片时使用——优先由主模型尝试处理图片,若主模型无法正常处理(报错/无法读取/无法识别/格式错误/内容块缺失等)则使用此技能进行视觉理解。此技能也应在用户要求生成图片/画图/配图时使用。

No reviews yet
0 installs
12 views
0.0% view→install

Install

$ agentstack add skill-hellowind777-hello-multimodal-hello-multimodal

✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access No
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

Are you the author of Hello Multimodal? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

HelloMultimodal

两个脚本均为 纯标准库,零外部依赖

路由

消息含图片 → 优先主模型处理
             ├── 成功 → 完成
             └── 无法处理 → 使用此技能视觉理解

要求生图 → 使用此技能图片生成

视觉理解通过 scripts/vision.py 调用外部多模态视觉 API,图片生成通过 scripts/generate.py 调用外部生图 API。

配置 (config.json)

{
  "channels": [
    {
      "name": "My Vision Channel",
      "base_url": "https://api.example.com",
      "api_key": "sk-your-api-key",
      "model": "gpt-4o",
      "image_model": "gpt-image-2",
      "image_api_key": "",
      "image_base_url": "",
      "responses_model": "",
      "chat_model": "",
      "vision": true,
      "generate": true,
      "priority": 1
    }
  ],
  "defaults": {
    "max_tokens": 4096,
    "timeout_seconds": 300,
    "retry_count": 2
  }
}

| 字段 | 说明 | |------|------| | model | 视觉理解模型 | | image_model | 生图模型(不填回退 model) | | image_api_key | 生图 API key(不填回退 api_key) | | image_base_url | 生图 base URL(不填回退 base_url) | | api_format | openai(默认)/ anthropic / sd-webui | | vision: true | 可用于视觉理解 | | generate: true | 可用于图片生成 | | priority | 越小越优先,失败自动 fallback |

详细模板见 config.example.json

工作流

> Windows 路径注意:bash 中反斜杠 \ 会被当作转义符吃掉。使用正斜杠 / 或单引号包裹:'C:/Users/xxx/img.png'

视觉理解

# 单图分析(默认 stdout,不产生文件)
python scripts/vision.py --image ./screenshot.png --prompt "描述图片内容"

# 批量分析
python scripts/vision.py --image-dir ./pages/ --prompt "批量分析"

# 指定渠道
python scripts/vision.py --channel 2 --image ./img.png --prompt "..."

图片生成

# 基本文生图(输出到 ./output/ 目录)
python scripts/generate.py --prompt "施工质量评分雷达图" --output ./output/chart.png

# 参考图编辑/变体
python scripts/generate.py --prompt "turn into oil painting" --image ./sketch.png --output ./output/painting.png

# 生成多张 / 指定渠道 / 调试
python scripts/generate.py --prompt "fantasy monster concept" --count 3 --output ./output/monster.png
python scripts/generate.py --channel 2 --prompt "..." --output ./output/img.png
python scripts/generate.py --prompt "test" --dry-run

> 视觉理解默认输出 stdout,不创建文件。图片生成输出到 ./output/,不写入桌面或用户目录。

规则

  • vision.py:按 priority 升序尝试渠道,失败自动下一个,不重试同一渠道
  • generate.py:每个端点按 retry_count 重试后切换到下一个端点,再切换到下一个渠道
  • 所有渠道失败后输出错误详情

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet — be the first.

Versions

  • v0.1.0 Imported from the upstream source.