name: chat-image-reader description: First universal image reader for 10+ chat platforms. Features: (1) Auto platform detection (Feishu, DingTalk, WeChat, Discord, Telegram, WhatsApp, Slack, Signal, LINE, iMessage), (2) Unified API with platform-specific download logic, (3) Comprehensive API reference with error codes and rate limits, (4) Fallback strategies for missing images. Triggers: "識別圖片", "讀取圖片", "analyze image", "read image from chat". No similar skill exists on ClawHub - fills a unique gap for cross-platform chat image analysis. author: Qingyu Zhu homepage: https://clawhub.ai metadata: openclaw: emoji: "🖼️" requires: bins: - python3 env: - FEISHU_APP_ID - FEISHU_APP_SECRET - DINGTALK_APP_KEY - DINGTALK_APP_SECRET - WECHAT_CORP_ID - WECHAT_CORP_SECRET - TELEGRAM_BOT_TOKEN - DISCORD_BOT_TOKEN envOptional: - DASHSCOPE_API_KEY - BAIDU_API_KEY - BAIDU_SECRET_KEY primaryEnv: FEISHU_APP_ID networkAccess: true permissions: - "Read chat messages" - "Download message attachments" - "Access platform APIs"

Chat Image Reader

7w4.net小蔥技能站收錄全網優質技能，值得收藏。

Universal image reading and analysis skill for multiple chat platforms.

Supported Platforms

Required Credentials

This skill requires API credentials to access chat platform messages. Configure at least one platform to use the skill.

Platform Credentials

Configuration Example

# Feishu (飛書) - Required for Feishu image reading
export FEISHU_APP_ID="cli_xxx"
export FEISHU_APP_SECRET="xxx"

# DingTalk (釘釘) - Required for DingTalk image reading
export DINGTALK_APP_KEY="dingxxx"
export DINGTALK_APP_SECRET="xxx"

# WeChat (企業微信) - Required for WeChat image reading
export WECHAT_CORP_ID="xxx"
export WECHAT_CORP_SECRET="xxx"

# Telegram - Required for Telegram image reading
export TELEGRAM_BOT_TOKEN="123456:ABC-xxx"

Security Notes

• Credentials are only used to download images from respective platforms
• No data is sent to external servers except the official platform APIs
• Images are stored temporarily in local temp directory
• Credentials should be configured via environment variables, not hardcoded

Workflow

Step 1: Detect Platform from Context

Check inbound_meta.channel or inbound_meta.provider to determine the chat platform:

{
  "channel": "feishu",      // Feishu
  "channel": "discord",     // Discord
  "channel": "telegram",    // Telegram
  "channel": "whatsapp",    // WhatsApp
  ...
}

Step 2: Get Image by Platform

Feishu (飛書)

1. Get message_id and image_key from message
2. Get tenant access token via API
3. Download from: GET /open-apis/im/v1/messages/{message_id}/resources/{image_key}
4. Save to temp file

DingTalk (釘釘)

1. Get downloadCode from message content
2. Get access token via API (appKey + appSecret)
3. Download from: GET /v1.0/robot/messageFiles/download?downloadCode={code}
4. Save to temp file

WeChat (微信/企業微信)

1. Get media_id from message
2. Get access token via API
3. Download from: GET https://qyapi.weixin.qq.com/cgi-bin/media/get?access_token={token}&media_id={id}
4. Save to temp file

Discord

1. Get attachment URL from message (Discord provides direct CDN URLs)
2. Download directly from URL
3. No authentication needed for public channels

Telegram

1. Get file_id from message
2. Call Bot API: GET https://api.telegram.org/bot{token}/getFile?file_id={file_id}
3. Download from: https://api.telegram.org/file/bot{token}/{file_path}

WhatsApp / Signal / Slack

1. Get media URL from message
2. Download directly (usually with token in header)

Local File Path

If user provides a local path directly: 1. Verify file exists 2. Use directly without download

Step 3: Analyze Image

Use the image tool with appropriate prompt:

image(
  image: "<local_path_or_url>",
  prompt: "描述圖片內容，包括文字、圖表、資料等關鍵資訊"
)

Platform-Specific Implementation

Feishu Implementation

# 1. Get tenant token
$body = @{ app_id = $appId; app_secret = $appSecret } | ConvertTo-Json
$token = (Invoke-RestMethod -Uri "https://open.feishu.cn/open-apis/auth/v3/tenant_access_token/internal" -Method Post -Body $body).tenant_access_token

# 2. Get message to find image_key
$message = Invoke-RestMethod -Uri "https://open.feishu.cn/open-apis/im/v1/messages/$messageId" -Headers @{ Authorization = "Bearer $token" }
$imageKey = $message.data.items[0].body.content | ConvertFrom-Json | Select-Object -ExpandProperty image_key

# 3. Download image
Invoke-WebRequest -Uri "https://open.feishu.cn/open-apis/im/v1/messages/$messageId/resources/$imageKey" -Headers @{ Authorization = "Bearer $token" } -OutFile $outputPath

DingTalk Implementation

# 1. Get access token
$body = @{ appKey = $appKey; appSecret = $appSecret } | ConvertTo-Json
$token = (Invoke-RestMethod -Uri "https://api.dingtalk.com/v1.0/oauth2/accessToken" -Method Post -Body $body).accessToken

# 2. Download image using downloadCode
# For robot messages:
Invoke-WebRequest -Uri "https://api.dingtalk.com/v1.0/robot/messageFiles/download?downloadCode=$downloadCode" -Headers @{ "x-acs-dingtalk-access-token" = $token } -OutFile $outputPath

# For user messages (via stream):
# Use conversation message download API

WeChat (企業微信) Implementation

# 1. Get access token
$token = (Invoke-RestMethod -Uri "https://qyapi.weixin.qq.com/cgi-bin/gettoken?corpid=$corpId&corpsecret=$corpSecret").access_token

# 2. Download media
Invoke-WebRequest -Uri "https://qyapi.weixin.qq.com/cgi-bin/media/get?access_token=$token&media_id=$mediaId" -OutFile $outputPath

Discord Implementation

# Discord attachments are direct URLs
$attachmentUrl = $message.attachments[0].url
Invoke-WebRequest -Uri $attachmentUrl -OutFile $outputPath

Telegram Implementation

# 1. Get file path from file_id
$fileInfo = Invoke-RestMethod -Uri "https://api.telegram.org/bot$token/getFile?file_id=$fileId"
$filePath = $fileInfo.result.file_path

# 2. Download file
Invoke-WebRequest -Uri "https://api.telegram.org/file/bot$token/$filePath" -OutFile $outputPath

Reply Context Handling

When user replies to an image message:

1. Check reply_to_id in inbound metadata
2. Fetch the original message using platform-specific API
3. Extract image from the original message
4. Proceed with download and analysis

Common Analysis Prompts

Document/Screenshot

識別圖片中的所有文字內容，保持原有格式和層次結構。

Chart/Table

分析圖片中的圖表或表格，提取所有資料和標籤。

UI Screenshot

描述介面佈局、功能按鈕、當前狀態等關鍵資訊。

Stock/Finance Chart

識別股票程式碼、價格、K線形態、成交量等財務資訊。

Job Posting

提取招聘資訊，包括職位名稱、職責、要求、薪資等關鍵內容。

General Image

描述圖片內容，包括主要元素、文字、資料等關鍵資訊。

Error Handling

Fallback Strategy

When automatic image retrieval fails:

1. Ask user to provide path: "請提供圖片的本地路徑"
2. Ask user to describe: "請描述圖片內容或複製圖片中的文字"
3. Request resend: "請重新發送圖片"

Temp File Management

• Save images to $workspace/temp_images/ or system temp
• Clean up after analysis (optional, for disk space)
• Use unique filenames with timestamp: img_{channel}_{timestamp}.jpg

Configuration Required

Feishu (飛書)

• FEISHU_APP_ID / channels.feishu.appId
• FEISHU_APP_SECRET / channels.feishu.appSecret
• Required scope: im:message:readonly, im:resource

DingTalk (釘釘)

• DINGTALK_APP_KEY / channels.dingtalk.appKey
• DINGTALK_APP_SECRET / channels.dingtalk.appSecret
• Required permissions: IMessage, Chat

WeChat (企業微信)

• WECHAT_CORP_ID / channels.wechat.corpId
• WECHAT_CORP_SECRET / channels.wechat.corpSecret
• Required permissions: media_get

Telegram

• Bot token configured in OpenClaw

Discord

• Bot token configured in OpenClaw

Notes

• Most platforms provide images in JPG/PNG format
• Large images may need resizing before analysis
• Consider rate limits for API calls
• Always validate user has permission to access the image