Can I use this for enterprise-grade detection?

Yes, it includes patterns for Microsoft Presidio, which uses Natural Language Processing (NLP) to identify complex entities that simple regex might miss.

How does it integrate with Langfuse?

The skill provides specific mask callback implementations that intercept data at the Langfuse client level, ensuring sensitive information is redacted before it is transmitted to your observability dashboard.

What is the PII Masking Patterns skill?

It is a specialized capability for Claude Code that provides production-ready code patterns to detect and redact sensitive information like emails, names, and credit card numbers from AI-generated logs and traces.

Does it support reversible anonymization?

Yes, the skill includes patterns for using LLM Guard with a Vault, which allows you to anonymize data for the LLM while safely deanonymizing it for your internal application logic.

Is this useful for GDPR compliance?

Absolutely. It helps developers adhere to 'privacy by design' principles by preventing the storage of PII in logging and monitoring systems.

PII Masking Patterns

Name: PII Masking Patterns
Author: yonatangross

byyonatangross

•

セキュリティとテスト

Protects sensitive user data in LLM observability pipelines by detecting and masking PII using regex, Microsoft Presidio, and LLM Guard.

PII Masking Patterns provides a comprehensive set of implementation strategies for securing LLM applications by identifying and redacting Personally Identifiable Information (PII) before it enters observability logs or third-party tracing tools. It offers ready-to-use patterns for standard regex masking, enterprise-grade NLP detection with Microsoft Presidio, and advanced sanitization through LLM Guard, ensuring your production AI pipelines remain compliant with GDPR and other data privacy regulations without sacrificing observability.

主な機能

01Seamless Langfuse integration via custom mask callbacks

02Pre-logging sanitization for structured logging with structlog and loguru

03Lightweight regex redaction for emails, phone numbers, and SSNs

04Reversible anonymization using LLM Guard and Vault patterns

05Automated PII detection using Microsoft Presidio and NLP models

0669 GitHub stars

ユースケース

01Implementing reversible anonymization to allow secure debugging of production LLM outputs

02Standardizing PII redaction across enterprise AI logging and monitoring stacks

03Masking customer identifiers in Langfuse traces to maintain GDPR and SOC2 compliance

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add yonatangross/orchestkit pii-masking-patterns

For use in Claude.ai and ChatGPT

Download Skill