Name: Ai Safety Auditor
Author: mdbabumiamssm

The AI Safety Auditor is a critical "human-in-the-loop" simulator and automated guardrail system. It intercepts outputs from other clinical agents to ensure they meet medical safety standards, do not contain Protected Health Information (PHI) where inappropriate, and are free from harmful hallucinations.

When to Use This Skill

As a final check before any clinical agent output is shown to a user.
To audit historical logs of agent interactions for compliance.
When detecting potential bias in diagnosis or treatment recommendations.
To verify that citations in a generated report actually exist (hallucination check).

Core Capabilities

PHI Scrubbing Verification: Ensures no identifiers leaked into non-secure outputs.
Hallucination Detection: Cross-references generated claims against trusted knowledge bases.
Bias Scanning: Checks for demographic or socioeconomic bias in clinical reasoning.
Contraindication Check: Verifies treatment recommendations against patient allergies/conditions.

When to Use This Skill

As a final check before any clinical agent output is shown to a user.
To audit historical logs of agent interactions for compliance.
When detecting potential bias in diagnosis or treatment recommendations.
To verify that citations in a generated report actually exist (hallucination check).

Core Capabilities

PHI Scrubbing Verification: Ensures no identifiers leaked into non-secure outputs.
Hallucination Detection: Cross-references generated claims against trusted knowledge bases.
Bias Scanning: Checks for demographic or socioeconomic bias in clinical reasoning.
Contraindication Check: Verifies treatment recommendations against patient allergies/conditions.

Ai Safety Auditor

When to Use This Skill

Core Capabilities

Ai Safety Auditor

When to Use This Skill

Core Capabilities

Workflow

Example Usage

Healthcare Cdss Patterns

Drug Discovery

Qmd

Attack Tree Construction

Azure Ai Anomalydetector Java

Viboscope