Overview

The pii-redaction skill detects personally identifiable information in conversational data and applies configurable redaction strategies to protect user privacy. It supports reversible anonymization, audit logging, and language-aware configuration to satisfy privacy and compliance needs.

How This Skill Works

It uses a PII detector to flag sensitive data across defined types, then applies a configured redaction method such as masking, pseudonymization, tokenization, or encryption. Redaction results are logged for compliance, and the system can be tailored to language support and custom patterns for organization-specific PII.

When to Use It

• Moderate user-generated chat content to redact emails, phone numbers, and names before storage or display.
• Configure system prompts and guardrails to prevent PII leakage to users or agents.
• Store or transmit chat logs for regulatory compliance with audit trails.
• Handle financial or health-related conversations by applying redaction or anonymization to sensitive fields.
• Test and validate redaction coverage with diverse data patterns and languages to ensure robustness.

Quick Start

1. Step 1: Define PII types to detect and set the detection sensitivity ( Names, Contact, Financial, Government IDs, Medical, Custom ).
2. Step 2: Choose a redaction method ( masking, pseudonymization, tokenization, or encryption ) and wire it into the redaction pipeline.
3. Step 3: Enable compliance logging and run tests with diverse data to verify coverage, reversibility where configured, and audit trails.

Best Practices

• Comprehensive PII coverage by enumerating all relevant types (Names, Contact, Financial, Government IDs, Medical, Custom).
• Regular pattern updates to keep detection effective against new formats and aliases.
• Audit logging for every redaction event and reversible anonymization activity.
• Compliance alignment with regional data protection laws and internal policies.
• Testing with diverse data, languages, and edge cases to ensure reliable performance.

Example Use Cases

• Redacting emails and phone numbers from customer support transcripts before analytics or storage.
• Masking names and usernames in content moderation logs to protect user identity.
• Pseudonymizing patient identifiers in medical chatbot conversations for research or support.
• Tokenizing or encrypting sensitive fields like account numbers in logs for secure storage.
• Defining and applying custom patterns for organization-specific PII such as employee IDs.

Frequently Asked Questions