intelligent-inspection

Intelligent workplace inspection system with guided setup, configurable inspection tasks, AI-powered image analysis, and Feishu alerting. Use when you need t...

Security Scan

The skill's declared purpose (camera capture → AI analysis → Feishu alerts) is plausible, but multiple inconsistencies and vague runtime instructions (notably where credentials/config are stored, differing config schemas, and use of exec/web_fetch without clear safeguards) make the package risky until those gaps are clarified.

Purpose & Capability

The requested capabilities (camera capture, AI image analysis, Feishu alerts) are consistent with the skill's stated purpose. The skill does not request unrelated cloud credentials or unusual binaries. However, there are inconsistencies across the manifest and files: SKILL.md says config is stored at ~/.openclaw/workspace/intelligent-inspection-config.json, example/config-template JSONs use different field names (camera.url vs apiUrl, ai vs inspection, alerts vs alert), and index.js actually writes config to a local skillDir/config.json. These mismatches mean the runtime behavior may differ from the documentation.

Instruction Scope

SKILL.md instructs the agent to capture images, call OpenClaw vision models, and send alerts via message/web_fetch/exec. Those instructions are within the apparent purpose but are vague about exact endpoints and retention: it claims 'no data retention' and 'images processed through secure AI endpoints' but provides no enforcement or deletion steps. The doc's mention of shell exec/web_fetch gives the agent the ability to run arbitrary commands or fetch arbitrary URLs during capture, which is expected for camera access but increases risk if prompts/inputs are not strictly validated. Also SKILL.md's declared config path differs from what index.js uses, so the code may read/write a different file than the docs promise.

Install Mechanism

This is an instruction-only skill with no install spec; no packages are pulled automatically and no archives are extracted. That minimizes installer risk.

Credentials

The skill reasonably requires camera API tokens and Feishu bot permissions for its function and declares no additional environment variables. But it stores credentials in a JSON config file — the docs say this will be in the user's ~/.openclaw workspace, while index.js currently saves to the skill directory (skillDir/config.json). Storing credentials in the package directory could be less secure (permission/visibility concerns). In addition, the different config schemas across files may cause credentials to be written or read under unexpected keys/locations, increasing the chance of accidental exposure.

Persistence & Privilege

always is false and the skill is user-invocable only. It does not request persistent system-wide privileges or modify other skills. Autonomous invocation is allowed by default (not flagged).

What to consider before installing

Before installing or running this skill: (1) Ask the author to reconcile the documentation and code: which config file path is authoritative (~/.openclaw/workspace/... vs skillDir/config.json) and which JSON schema is expected (camera.url vs apiUrl, inspection vs ai, alerts vs alert). (2) Verify where camera credentials and Feishu tokens will be stored and prefer a secure user-owned location (home directory) with proper file permissions; do not store secrets in system/package directories. (3) Confirm the exact AI endpoints used and whether images are deleted after processing; request explicit deletion/retention logic if you require no retention. (4) Inspect or sandbox the skill runtime (or run it with least privilege) because SKILL.md allows exec/web_fetch; ensure inputs used for any shell or fetch calls are validated and cannot be influenced by untrusted text. (5) Test in a controlled environment with dummy credentials and dummy cameras before deploying to production. If the author cannot clarify these inconsistencies, treat the skill as risky and avoid giving it real camera or Feishu credentials.

Intelligent Inspection - 智能巡检

Overview

Intelligent Inspection is an intelligent workplace monitoring system that:

• Captures images from surveillance cameras
• Uses AI to analyze images based on user-defined inspection criteria
• Sends alerts via Feishu (or other configured channels)
• Supports guided configuration and isolated settings

This skill is designed for workplace safety, security monitoring, and employee presence verification.

Core Features

1. Guided Configuration

• First-time setup wizard collects camera parameters
• User-friendly prompts for API endpoints, tokens, and device settings
• Configuration isolation in dedicated config file

2. Flexible Inspection Tasks

• Customizable AI prompts for different inspection scenarios
• Support for multiple camera devices and channels
• Configurable patrol schedules and triggers

3. AI-Powered Analysis

• Integration with OpenClaw's AI models
• User-defined analysis criteria through custom prompts
• Structured response handling for consistent results

4. Multi-Channel Alerting

• Primary integration with Feishu messaging
• Automatic fallback to other configured channels
• Rich message formatting with images and metadata

Setup Requirements

Camera System Requirements

• Supported camera APIs:
  • EZVIZ Cloud API (萤石云)
  • Generic RTSP/ONVIF cameras (via camsnap skill)
  • Custom HTTP-based capture endpoints

Authentication Requirements

• Camera API access tokens or credentials
• Feishu bot permissions (if using Feishu alerts)

OpenClaw Requirements

• feishu channel configured (for Feishu alerts)
• AI model with vision capabilities (e.g., qwen3-max, GPT-4 Vision)

Usage Scenarios

Employee Presence Monitoring

• Detect if employees are at their workstations
• Monitor break room usage
• Track attendance during working hours

Security Patrols

• Verify door/window status
• Check for unauthorized access
• Monitor restricted areas

Facility Inspections

• Equipment status verification
• Environmental condition monitoring
• Compliance checks

Workflow

1. Initial Setup (First Run)

When the skill runs for the first time:

1. Prompt user for camera system type
2. Collect API endpoint, access token, device serial, channel number
3. Ask for default inspection prompt template
4. Confirm alert channel preferences
5. Save configuration to ~/.openclaw/workspace/intelligent-inspection-config.json

2. Patrol Execution

For subsequent runs:

1. Load configuration from config file
2. Capture image using configured camera parameters
3. Generate AI analysis prompt based on patrol task
4. Execute AI analysis using OpenClaw's vision model
5. Format and send alert via configured channels

3. Configuration Management

• Configuration stored in isolated JSON file
• Users can edit config file directly for advanced settings
• Skill validates configuration on each run
• Option to reset configuration and re-run setup

Configuration File Structure

The configuration file (intelligent-inspection-config.json) contains:

{
  "camera": {
    "type": "ezviz",
    "apiUrl": "https://open.ys7.com/api/open/cloud/v1/capture/save",
    "accessToken": "your-access-token",
    "deviceSerial": "C12345678",
    "channelNo": "1",
    "projectId": "intelligent-inspection-project"
  },
  "inspection": {
    "defaultPrompt": "请分析这张图片中是否有员工在工位上。如果没有人，请回复'离岗'；如果有人，请回复'在岗'。",
    "alertOn": ["离岗"],
    "includeImage": true
  },
  "alerts": {
    "enabled": true,
    "channels": ["feishu"],
    "fallbackToDefault": true
  }
}

Error Handling

• Invalid camera credentials: Clear error messages with setup guidance
• AI analysis failures: Retry logic with fallback prompts
• Alert delivery failures: Log errors and attempt alternative channels
• Missing configuration: Automatically trigger setup wizard

Privacy and Security

• Camera credentials stored locally only
• Images processed through secure AI endpoints
• No data retention beyond immediate patrol execution
• Compliance with workplace monitoring regulations recommended

Integration Points

OpenClaw Tools Used

• message: For sending alerts via configured channels
• web_fetch/exec: For camera image capture
• AI model calls: For image analysis

External Dependencies

• Camera system APIs (EZVIZ, RTSP, etc.)
• Feishu bot (if configured as alert channel)

Examples

Basic Employee Monitoring

执行智能巡检任务

Custom Inspection Task

执行智能巡检：检查会议室是否有人使用

Reset Configuration

重置智能巡检配置

Version History

• v1.0.0: Initial release with EZVIZ support and Feishu alerts