Can I use this for PII detection in code?

Yes, the 2025 enhancements include specific support for detecting PII within code syntax, comments, and variable names.

How does contextual grounding help my application?

It validates the factual accuracy and relevance of AI responses against your source data, which is critical for preventing hallucinations in RAG (Retrieval-Augmented Generation) applications.

What is Automated Reasoning in this context?

It is a policy verification feature that uses mathematical logic to ensure model outputs strictly adhere to your defined rules with up to 99% accuracy.

Does this skill only work with Amazon Bedrock models?

No, it uses the ApplyGuardrail API which is model-agnostic and can validate inputs and outputs for any LLM, including OpenAI and Google Gemini.

Amazon Bedrock Guardrails

Name: Amazon Bedrock Guardrails
Author: adaptationio

byadaptationio

Security & Testing

Implements comprehensive AI safety policies and content moderation using Amazon Bedrock's multi-layered safeguard framework.

About

This skill provides a robust implementation of Amazon Bedrock Guardrails, allowing developers to secure generative AI applications through six distinct safeguard policies. It enables model-agnostic protection—working across Bedrock, OpenAI, and Gemini—by providing capabilities for content filtering, PII redaction, topic denial, custom word lists, contextual grounding for RAG applications, and automated reasoning with high mathematical accuracy. It is particularly useful for enterprise applications requiring strict compliance, hallucination prevention, and protection against prompt injection attacks.

Key Features

Automated reasoning for formal policy verification with 99% accuracy
Multi-layered content filtering for hate, violence, and prompt injection attacks
Automatic PII detection and redaction for SSNs, emails, and custom regex patterns
Topic denial to enforce domain-specific compliance and prevent unauthorized advice
0 GitHub stars
Contextual grounding to detect and prevent hallucinations in RAG systems

Use Cases

Implementing unified safety guardrails across multi-model AI architectures
Securing enterprise RAG applications against hallucinations and factual inaccuracies
Automating PII redaction and compliance in customer-facing AI chatbots

About

Key Features

Automated reasoning for formal policy verification with 99% accuracy
Multi-layered content filtering for hate, violence, and prompt injection attacks
Automatic PII detection and redaction for SSNs, emails, and custom regex patterns
Topic denial to enforce domain-specific compliance and prevent unauthorized advice
0 GitHub stars
Contextual grounding to detect and prevent hallucinations in RAG systems

Use Cases

Implementing unified safety guardrails across multi-model AI architectures
Securing enterprise RAG applications against hallucinations and factual inaccuracies
Automating PII redaction and compliance in customer-facing AI chatbots