imageReader

Reads and analyzes images from messages across 10+ chat platforms using platform-specific APIs and unified image processing.

Security Scan

The skill's code and runtime instructions match its stated purpose (downloading images from many chat platforms) but the registry metadata, SKILL.md, and the included script disagree about required credentials and environment variables — this mismatch and the broad set of platform tokens requested warrant caution before installing.

Purpose & Capability

The skill's behavior (download images from many chat platforms) is coherent with its name and description. However, the registry-level metadata claims no required environment variables or primary credential, while SKILL.md and the included Python downloader clearly require multiple sensitive platform credentials (FEISHU, DINGTALK, WECHAT, TELEGRAM, DISCORD, WhatsApp, Slack, LINE, etc.). This inconsistency between declared registry requirements and the actual runtime needs is a red flag (could be sloppy packaging or an attempt to hide sensitive requirements).

Instruction Scope

SKILL.md and scripts limit actions to detecting platform, calling official platform APIs, saving downloaded images to a temp file, and invoking an image-analysis tool. There are no obvious instructions to read unrelated files or exfiltrate data to arbitrary endpoints. However: (1) SKILL.md asserts 'No data is sent to external servers except the official platform APIs' but the skill references optional external OCR/API keys (Baidu/Dashscope) and the downstream 'image(...)' tool — the actual destination of image analysis depends on the runtime image tool and could involve sending image data to an external model or service. (2) The skill instructs accessing inbound_meta and message contents (expected) but that grants broad read access to chat content.

Install Mechanism

This is instruction-only with a bundled Python script and no install spec that downloads arbitrary archives. No third-party install URLs, package installs, or extract-from-URL steps are present. The included script will be available on disk when the skill is installed; it uses the requests library but no additional package installation is specified in an install step.

Credentials

Functionally, the skill legitimately needs platform-specific tokens to download messages/images. But the manifest is inconsistent: top-level registry metadata lists no env vars, SKILL.md metadata lists a subset (python3 + FEISHU/DISCORD/TELEGRAM), and the Python script expects many more environment variables (WECHAT, DINGTALK, SLACK, WHATSAPP tokens, etc.) and/or tokens passed as CLI args. Asking for many platform credentials is proportionate to a multi-platform downloader only if declared transparently; the current mismatch (and the requirement for multiple powerful tokens that can read message content) increases risk if users supply full-privilege credentials.

Persistence & Privilege

The skill does not request always:true and does not claim to modify other skills or global configs. disable-model-invocation is false (normal). Note: autonomous invocation combined with broad platform credentials would enlarge the blast radius, but autonomous invocation alone is the platform default and not a standalone concern.

What to consider before installing

Don't install blindly. Key points to consider before proceeding: - Inconsistency: The registry metadata claims no required env vars, but SKILL.md and the included Python script require many sensitive platform tokens. Ask the author to reconcile the manifest and provide a clear, minimal list of required env vars. - Sensitive credentials: This skill needs tokens that can read chat messages; only provide tokens with the minimum scopes (read-only, narrow workspace/chat scopes) and prefer test accounts or sandbox credentials first. Rotate tokens after testing. - Data flow: Confirm where images are sent for analysis. The SKILL.md's image(...) call may forward image data to the model provider or third-party services; demand explicit data-handling/retention details and logs of network destinations. - Audit the code: The included download_image.py appears to only call official platform APIs and write files to temp directories, but verify there are no hidden endpoints, telemetry, or additional network calls in the full script. Run it in an isolated environment (sandbox/VM) and inspect network traffic while testing. - Origin and maintenance: The skill owner is unknown and there's no homepage; prefer skills from known authors or ask for contact/maintainer info and a verifiable repository. - Least privilege & testing: If you must try it, limit scope (enable only one platform at a time), use scoped/test tokens, and monitor activity. If the author can't clarify the manifest mismatches and data handling, treat this as untrusted.

Chat Image Reader

Universal image reading and analysis skill for multiple chat platforms.

Supported Platforms

Platform	Channel	Image Source	Download Method
Feishu (飞书)	feishu	Message image_key	API + Tenant Token
DingTalk (钉钉)	dingtalk	Message downloadCode	API + Access Token
WeChat (微信)	wechat	Media file_id	API + Access Token
Discord	discord	Attachment URL	Direct download
Telegram	telegram	file_id	Bot API
WhatsApp	whatsapp	Media URL	Direct download
Signal	signal	Attachment	Direct download
Slack	slack	File URL	Direct download
iMessage	imessage	Attachment path	Local file
LINE	line	Message content	API

Required Credentials

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

Platform Credentials

Platform	Required Environment Variables	Notes
Feishu (飞书)	FEISHU_APP_ID, FEISHU_APP_SECRET	Required scopes: im:message:readonly, im:resource
DingTalk (钉钉)	DINGTALK_APP_KEY, DINGTALK_APP_SECRET	Required permissions: IMessage, Chat
WeChat (企业微信)	WECHAT_CORP_ID, WECHAT_CORP_SECRET	Required permissions: media_get
Telegram	TELEGRAM_BOT_TOKEN	From @BotFather
Discord	DISCORD_BOT_TOKEN	From Discord Developer Portal
WhatsApp	WHATSAPP_TOKEN	From Meta Business API

Configuration Example

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

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

# WeChat (企业微信) - 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 (飞书)

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 (钉钉)

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/企业微信)

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: "描述图片内容，包括文字、图表、数据等关键信息"
)

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 (企业微信) 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

识别图片中的所有文字内容，保持原有格式和层次结构。

Chart/Table

分析图片中的图表或表格，提取所有数据和标签。

UI Screenshot

描述界面布局、功能按钮、当前状态等关键信息。

Stock/Finance Chart

识别股票代码、价格、K线形态、成交量等财务信息。

Job Posting

提取招聘信息，包括职位名称、职责、要求、薪资等关键内容。

General Image

描述图片内容，包括主要元素、文字、数据等关键信息。

Error Handling

Error	Cause	Solution
Image not found	No image in message	Ask user to resend or provide path
Download failed	Permission/Network issue	Check permissions, retry
API error	Token expired/invalid	Refresh token and retry
Analysis failed	Image unclear/unsupported	Try alternative prompt

Fallback Strategy

When automatic image retrieval fails:

1. Ask user to provide path: "请提供图片的本地路径"
2. Ask user to describe: "请描述图片内容或复制图片中的文字"
3. Request resend: "请重新发送图片"

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 (飞书)

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

DingTalk (钉钉)

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

WeChat (企业微信)

• 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