Skills Azure AI Content Safety SDK for Python
📦

Azure AI Content Safety SDK for Python

Safe

Detect harmful content with Azure AI Content Safety

User-generated and AI-generated content can contain harmful material that violates community guidelines. This skill helps you automatically detect hate speech, self-harm content, sexual content, and violence with configurable severity thresholds.

Supports: Claude Codex Code(CC)
📊 69 Adequate
1

Download the skill ZIP

2

Upload in Claude

Go to Settings → Capabilities → Skills → Upload skill

3

Toggle on and start using

Test it

Using "Azure AI Content Safety SDK for Python". Text containing mild profanity but no hate speech

Expected outcome:

Hate: severity 0 (Safe), SelfHarm: severity 0 (Safe), Sexual: severity 0 (Safe), Violence: severity 0 (Safe)

Using "Azure AI Content Safety SDK for Python". Image containing weapons or violent imagery

Expected outcome:

Violence: severity 4 (Medium), other categories: severity 0 (Safe). Content flagged for review based on threshold settings.

Using "Azure AI Content Safety SDK for Python". Text matching custom blocklist term with halt_on_blocklist_hit enabled

Expected outcome:

BlocklistsMatch: true. Blocked term: [TERM]. Analysis halted before AI processing. Action: reject submission.

Security Audit

Safe
v1 • 2/24/2026

This skill contains only documentation for the Azure AI Content Safety Python SDK. Static analysis scanned 0 files with 0 security issues detected. The skill provides legitimate instructions for content moderation using Azure services with proper credential handling via environment variables. No executable code or malicious patterns found.

0
Files scanned
0
Lines analyzed
0
findings
1
Total audits
No security issues found
Audited by: claude

Quality Score

38
Architecture
100
Maintainability
87
Content
31
Community
100
Security
74
Spec Compliance

What You Can Build

Social Media Content Moderation

Automatically screen user posts and comments for harmful content before publishing to your platform.

AI Output Safety Filtering

Pre-screen AI-generated responses to ensure they meet safety guidelines before displaying to end users.

Chat Application Safety

Real-time monitoring of chat messages to detect and block harmful content in customer support or community chats.

Try These Prompts

Basic Text Analysis 
Analyze this text for harmful content: [INSERT TEXT]. Report severity levels for hate, self-harm, sexual, and violence categories.
Custom Blocklist Setup 
Create a blocklist named [BLOCKLIST_NAME] for my domain. Add these terms: [TERM1, TERM2, TERM3]. Configure text analysis to halt when blocked terms are detected.
Image Content Analysis 
Analyze this image for harmful visual content: [IMAGE_FILE or URL]. Use 8-severity scale and return results for all harm categories.
Multi-Stage Content Pipeline 
Build a content moderation pipeline that: 1) Checks text against custom blocklists first, 2) Analyzes remaining content with Azure AI, 3) Applies different actions based on severity thresholds (0-2: allow, 4: flag for review, 6+: block). Log all results for audit.

Best Practices

  • Set severity thresholds appropriate for your audience and platform guidelines
  • Use custom blocklists for domain-specific terms that Azure AI may not recognize
  • Log all analysis results for audit trails and continuous improvement of moderation policies

Avoid

  • Do not rely solely on automated moderation for high-stakes content decisions
  • Avoid using default severity thresholds without testing against your specific content types
  • Do not store or log raw harmful content in analysis results for compliance reasons

Frequently Asked Questions

What Azure subscription do I need for Content Safety?
You need an Azure subscription with a Content Safety resource provisioned. The service is billed per 1000 text records or images analyzed.
How do I choose between 4-level and 8-level severity?
Use 4-level (0, 2, 4, 6) for simple allow/flag/block workflows. Use 8-level (0-7) when you need finer control over moderation thresholds.
Can I analyze content in multiple languages?
Yes, Azure AI Content Safety supports multiple languages. Check Azure documentation for the current list of supported languages.
What is the difference between blocklists and AI analysis?
Blocklists check for exact term matches you define. AI analysis detects harmful concepts and context that may not contain specific keywords.
How do I handle false positives in content moderation?
Implement a human review workflow for flagged content. Adjust severity thresholds and refine blocklists based on review outcomes.
Is API Key or Entra ID authentication more secure?
Entra ID is recommended for production environments as it supports managed identities and avoids storing secrets. API keys are simpler for development.

Developer Details

Author

sickn33

License

MIT

Repository

https://github.com/sickn33/antigravity-awesome-skills/tree/main/skills/azure-ai-contentsafety-py

Ref

main

File structure

📄 SKILL.md