security-review-owasp-secure-ai-model-ops

What problem does it solve?

This Skill helps reviewers find security weaknesses in AI model operations, especially where prompts, tools, retrieval, memory, or external services can be abused to cause unsafe behavior or leak sensitive context.

Core Features & Use Cases

• Trust Boundary Analysis: Traces how untrusted input can influence model instructions, tool calls, and downstream actions.
• Abuse Path Review: Focuses on prompt injection, unsafe automation, secret exposure, and cross-boundary data leakage.
• Operational Security Checks: Reviews logging, monitoring, governance, and runtime controls for AI-driven workflows.
• Use Case: A reviewer can use this Skill to assess an AI agent that reads documents and calls tools, then verify whether user content can override policy or trigger privileged actions.

Quick Start

Use the security-review-owasp-secure-ai-model-ops skill to review the AI agent flow for prompt injection, unsafe tool execution, and context leakage.