skills.rest
microsoft
microsoftOFFICIAL
Explore all skills by this author →
Website𝕏@OpenAtMicrosoftGitHub

owasp-agentic

Identifies, assesses, and remediates AI agent security risks using the OWASP Agentic Top 10.

1.4k|284|Updated Nov 2, 2025
Search Tags:#owasp#vulnerability assessment#remediation#threat detection#prompt injection#agentic security#AI Agents
One-click install
npx skills add https://github.com/microsoft/hve-core --skill owasp-agentic
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: owasp-agentic
Source: https://github.com/microsoft/hve-core/tree/main/.github/skills/security/owasp-agentic
Command: npx skills add https://github.com/microsoft/hve-core --skill owasp-agentic

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

AI agent systems introduce new security risks such as goal hijacking, tool misuse, and memory poisoning that traditional security guidance does not cover. This Skill provides a structured, machine-readable knowledge base of the OWASP Top 10 for Agentic Applications (2026) so agents and developers can identify, assess, and remediate these vulnerabilities.

Core Features & Use Cases

  • Vulnerability Reference Library: Ten normative documents (ASI01–ASI10) covering goal hijack, tool misuse, identity abuse, supply chain risks, code execution, memory poisoning, inter-agent communication, cascading failures, trust exploitation, and rogue agents.
  • Consistent Assessment Structure: Each reference includes description, risk, vulnerability checklist, prevention controls, attack scenarios, detection guidance, and remediation steps.
  • Use Case: While reviewing a multi-agent workflow design, query the skill to check whether inter-agent messages lack authentication, then apply the ASI07 prevention controls such as mutual TLS and signed agent cards.

Quick Start

Use the owasp-agentic skill to assess my agent architecture against the OWASP Agentic Top 10 and list applicable prevention controls.

Frequently Asked Questions about owasp-agentic

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

FAQPage Schema
How do I assess AI agent security risks with the OWASP Agentic Top 10?▼

Start with the vulnerability index to map your architecture to the ten ASI categories, then read each relevant reference document. Each one provides a vulnerability checklist, prevention controls, detection guidance, and remediation steps tailored to agentic systems.

What vulnerabilities does the OWASP Agentic Top 10 cover?▼

It covers ten risks: agent goal hijack, tool misuse, identity and privilege abuse, supply chain vulnerabilities, unexpected code execution, memory and context poisoning, insecure inter-agent communication, cascading failures, human-agent trust exploitation, and rogue agents.

How is agent goal hijack different from memory poisoning?▼

Goal hijack (ASI01) is direct manipulation of an agent's objectives or decision pathways, while memory poisoning (ASI06) corrupts stored context that biases future reasoning. Memory poisoning frequently leads to goal hijacking over time.

Does the OWASP Agentic Top 10 apply to multi-agent systems?▼

Yes, several entries specifically address multi-agent risks, including insecure inter-agent communication (ASI07), cascading failures across agents (ASI08), and cross-agent trust exploitation under identity abuse (ASI03).

What are the limitations of this OWASP agentic security reference?▼

It is a knowledge base of guidance, not an automated scanner, so it cannot detect vulnerabilities in running systems by itself. It also does not replace the OWASP LLM Top 10, which covers foundational model-level risks.

Related Skills

Explore More Software Engineering Skills→
Forexgod21Forexgod21
6

delegation-network-mapper

Map delegation chains and trust boundaries in multi-agent AI systems.

Community
Advanced
spinlockdevelopmentspinlockdevelopment
1

hardened-shell

Run agents inside an isolated Docker sandbox with read-only host mounts.

Official
Advanced
grant-vinegrant-vine
1

security-analyst

Identify and prioritize OWASP Top 10 vulnerabilities in web applications.

Community
Advanced
oakossoakoss
14

secure-ai

Implement defense-in-depth security for LLM pipelines and AI integrations.

Official
Advanced