llm-risk-assess

Assess LLM application security risks against the OWASP Top 10 for LLMs 2025.

141|14|Updated Mar 22, 2026
One-click install
npx skills add https://github.com/OWASP/secure-agent-playbook --skill llm-risk-assess-owasp
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: llm-risk-assess
Source: https://github.com/OWASP/secure-agent-playbook/tree/main/skills/llm-risk-assess
Command: npx skills add https://github.com/OWASP/secure-agent-playbook --skill llm-risk-assess-owasp

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Comprehensive evaluation for LLM-based applications against the OWASP Top 10 for LLMs 2025, enabling teams to identify, assess, and mitigate risks from prompt injection to data handling and supply chain issues.

Core Features & Use Cases

  • Threat modeling & architecture mapping: Document LLM provider choices, data flows, trust boundaries, and attack surfaces for end-to-end systems.
  • Automated security testing: Run prompt injection probes, test output handling, scan for secrets, and validate vector database and plugin trust.
  • OWASP-aligned reporting: Produce risk matrices, evidenced findings, remediation steps, and cross-reference to OWASP references for compliance.

Quick Start

Analyze your LLM-enabled application by outlining its data flow and threat model, then run the full assessment to produce your first risk report.

Frequently Asked Questions about llm-risk-assess

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

FAQPage Schema
How do I assess LLM application security risks using the OWASP Top 10 for LLMs 2025?

You can assess LLM application security risks by mapping data flows, threat modeling LLM providers, and running automated prompt injection probes to produce an OWASP-aligned risk matrix with remediation steps. The evaluation covers RAG components, tool integrations, and memory systems.

What is the best way to test for prompt injection and data poisoning in GenAI features?

Testing for prompt injection and data poisoning in GenAI features involves executing automated security probes and vulnerability checks across your architecture. The process validates vector database trust, tests output handling, and scans for secrets to identify system vulnerabilities.

Can I use automated red-team testing for LLM-enabled chatbots and agents?

Yes, you can use automated red-team testing for LLM-enabled chatbots and agents. The assessment runs targeted prompt injection probes, validates plugin trust boundaries, and checks memory systems to document the attack surface and identify security risks.

How do I map threat models and trust boundaries for RAG architectures?

To map threat models and trust boundaries for RAG architectures, you document LLM provider choices, data flows, and attack surfaces. This end-to-end architecture mapping identifies trust boundaries across vector databases, tool integrations, and memory systems.

Does the LLM security assessment generate compliance reports with remediation guidance?

Yes, the LLM security assessment generates compliance reports with remediation guidance. It produces risk matrices, evidenced findings, and remediation steps that cross-reference OWASP references for supply chain issues and data handling vulnerabilities.