Review LLM applications for OWASP Top 10 security weaknesses.
npx skills add https://github.com/UnitOneAI/SecuritySkills --skill llm-top-10-unitoneai
Please help me install this Agent Skill. Skill: llm-top-10 Source: https://github.com/UnitOneAI/SecuritySkills/tree/main/skills/ai-security/llm-top-10 Command: npx skills add https://github.com/UnitOneAI/SecuritySkills --skill llm-top-10-unitoneai
This Skill helps you produce reliable, framework-mapped security reviews for LLM-powered applications instead of vague or inconsistent guidance.
Review the target LLM codebase directory (or a specific file) by invoking llm-top-10 with the path you want assessed.
High-intent search queries and answers about installing and using this skill.
To audit an LLM application for OWASP Top 10 vulnerabilities, you review the target codebase by providing the directory path. The skill analyzes LLM API integrations, RAG pipelines, and tool-calling architectures to produce structured LLM01-LLM10 findings with severity ratings and remediation guidance.
RAG security involves protecting retrieval-augmented generation pipelines from weaknesses like prompt injection and vector/embedding poisoning. This skill evaluates RAG workflows to identify disclosure, supply chain risks, and output handling issues mapped to the OWASP LLM Top 10.
Yes, you can check conversational AI and tool calling code for excessive agency. The skill inspects agent and function-calling architectures to identify unbounded consumption, prompt leakage, and excessive permissions, generating prioritized remediation steps based on CWE mappings.
Yes, this security review covers model serving and fine-tuning flows. It evaluates these workflows for supply chain vulnerabilities, data poisoning, and misinformation risks, delivering an actionable findings report aligned with the 2025 OWASP Top 10 for Large Language Model Applications.
The best way to map LLM security findings to CWE identifiers is to perform a structured architectural and code inspection. This skill automatically generates LLM01-LLM10 findings mapped to CWE references, complete with severity ratings and prioritized remediation guidance.
To secure vector database embeddings against poisoning attacks, you analyze the embedding database workflows within your LLM application. The skill inspects these pipelines for vector and embedding weaknesses, output handling flaws, and disclosure risks mapped to OWASP LLM Top 10 categories.
0xPlaygroundsBuild LLM-powered Rust applications with a unified API for providers and vector stores.
mastepanoskiIdentify security vulnerabilities in LLM applications using the OWASP Top 10 for LLM Apps 2025.
kinnerkarmanishDetect vulnerabilities and misconfigurations across codebases using static analysis, dependency checks, and secret detection.
MuvonAudit LLM injection surfaces and produce OWASP-aligned defense recommendations.
