llm-top-10

Analyzes LLM applications for vulnerabilities across the ten OWASP AI categories.

Updated Apr 19, 2026

Search Tags:#security #vulnerability #llm #appsec #ml #vector-store #prompt-injection

Ready to run

npx skills add https://github.com/do360now/security-agents --skill llm-top-10

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: llm-top-10
Source: https://github.com/do360now/security-agents/tree/main/.claude/skills/llm-top-10
Command: npx skills add https://github.com/do360now/security-agents --skill llm-top-10

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill enables systematic security evaluation of AI applications leveraging large language models, focusing on identifying vulnerabilities and risks.

Core Features & Use Cases

Risk Assessment: Checks LLM codebases for OWASP Top 10 vulnerabilities specific to AI and LLM systems.
Automated Analysis: Examines prompt construction, data handling, vector store security, and output sanitization.
Use Case: Security engineers can run this Skill to generate comprehensive reports on the security posture of an AI-powered product, including specific mitigation steps.

Quick Start

Review the SKILL.md to understand the ten OWASP categories, then execute analysis routines on your codebase or infrastructure setup to identify gaps.

Frequently Asked Questions about llm-top-10

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I scan my LLM application for prompt injection vulnerabilities?▼

To identify data leakage in LLM systems, perform automated security analysis on data handling and vector store configurations. This review maps data flows against OWASP Top 10 categories to detect leakage paths and enforce best practices for sensitive information protection.

What is the best way to check vector store security in AI applications?▼

Checking vector store security in AI applications involves examining infrastructure configurations for poisoning risks and unauthorized access. A systematic review identifies vulnerabilities in vector store integration and supply chain dependencies, ensuring robust data protection.

Does an automated security review cover OWASP Top 10 risks for LLM systems?▼

Yes, an automated security review covers OWASP Top 10 risks for LLM systems. It systematically evaluates codebase, architecture, and dependencies to detect vulnerabilities like prompt injection, data leakage, and output misuse while enforcing best practices.

How do I perform a security audit on my AI codebase for output misuse?▼

Perform a security audit on your AI codebase for output misuse by analyzing sanitization routines and application logic. This review identifies potential output manipulation vulnerabilities and provides actionable remediation insights to secure LLM responses.

Can I assess dependency safety and supply chain risks for ML models?▼

Yes, you can assess dependency safety and supply chain risks for ML models. The review evaluates external libraries and model sources against OWASP criteria to detect poisoning and supply chain vulnerabilities, ensuring secure integration.

Why does my LLM application need an architecture security evaluation?▼

Your LLM application needs an architecture security evaluation to uncover systemic risks across data handling and model interactions. Reviewing architecture identifies gaps in configurations and enforces best practices, mitigating vulnerabilities before deployment.