Does it support image analysis as well as text?

Yes, it can analyze images provided as base64 encoded strings from local files or via direct image URLs.

Can I use custom banned words with this skill?

Yes, the skill includes support for the BlocklistClient, allowing you to create, update, and analyze content against custom text blocklists for domain-specific needs.

What categories of harm can this skill detect?

The skill detects four primary categories: Hate, Sexual, Violence, and Self-Harm, providing a severity score for each to help you decide on moderation actions.

How detailed are the severity scores provided by the SDK?

By default, the SDK returns four severity levels (0, 2, 4, 6), but it can be configured to provide an eight-level scale (0-7) for more granular control.

What are the authentication requirements?

The skill supports both Azure Key Credential (API Key) and Microsoft Entra ID (DefaultAzureCredential) for secure access to Azure AI services.

Azure AI Content Safety for Python

Name: Azure AI Content Safety for Python
Author: sickn33

bysickn33

•

31,721

•

데이터 과학 및 ML

Detects and filters harmful text and image content using Azure's AI-driven moderation and multi-severity classification.

This skill integrates the Azure AI Content Safety Python SDK into the Claude Code environment, enabling developers to build robust content moderation systems. It provides standardized implementation patterns for analyzing text and images for categories such as hate speech, self-harm, sexual content, and violence. By leveraging multi-severity scoring and customizable blocklists, it allows for fine-grained control over user-generated or AI-generated content, making it an essential tool for applications requiring high safety standards and compliance with moderation policies.

주요 기능

01Custom blocklist management for domain-specific term filtering

02Multi-category text analysis for Hate, Self-Harm, Sexual, and Violence

03Dual authentication support for Azure API Key and Entra ID

04Adjustable severity scales with 4-level or 8-level granularity

05Image moderation via local file upload or public blob URLs

0631,721 GitHub stars

사용 사례

01Managing custom lists of prohibited terms for niche online communities

02Pre-filtering AI model outputs to ensure safety and compliance

03Moderating user-generated comments or profiles in real-time

주요 기능

01Custom blocklist management for domain-specific term filtering

02Multi-category text analysis for Hate, Self-Harm, Sexual, and Violence

03Dual authentication support for Azure API Key and Entra ID

04Adjustable severity scales with 4-level or 8-level granularity

05Image moderation via local file upload or public blob URLs

0631,721 GitHub stars

사용 사례

01Managing custom lists of prohibited terms for niche online communities

02Pre-filtering AI model outputs to ensure safety and compliance

03Moderating user-generated comments or profiles in real-time