Organize LLM application security guidelines for building, reviewing, and deploying AI systems.
npx skills add https://github.com/semgrep/skills --skill llm-security
Please help me install this Agent Skill. Skill: llm-security Source: https://github.com/semgrep/skills/tree/main/skills/llm-security Command: npx skills add https://github.com/semgrep/skills --skill llm-security
This Skill provides a comprehensive framework of security guidelines for building, reviewing, and deploying LLM-powered applications, reducing risk across the development lifecycle.
Review and apply the OWASP Top 10 for LLM Applications 2025 guidelines to your next AI project.
High-intent search queries and answers about installing and using this skill.
To secure LLM applications against prompt injection and data leakage, apply defense-in-depth guardrails that define concrete patterns for sensitive data handling, input validation, and robust output handling across your model pipelines.
Security guidelines for RAG and AI model pipelines should enforce access control, rate limiting, and data poisoning prevention to ensure supply chain integrity and standardized risk monitoring across different environments.
Prevent data poisoning in LLM supply chains by implementing supply chain integrity guardrails that standardize security reviews, enforce auditing, and continuously monitor risks during the deployment of AI systems.
Yes, you can use this framework to standardize LLM security reviews across multiple teams by applying a structured playbook that enforces consistent access control, output validation, and risk monitoring for all projects.
The best way to implement defense-in-depth for LLM apps is to establish concrete guardrails across the development lifecycle, covering prompt injection, supply chain integrity, and output validation for production workflows.
Yes, the OWASP Top 10 for LLM Applications 2025 guidelines cover access control and auditing, providing actionable patterns and secure implementations to mitigate risks in production AI workflows.
ouakarDocument production AI/ML stack usage, LLM calls, pipelines, and cost drivers.
vKenjoConfigure authentication security settings including rate limiting, secrets, OAuth encryption, and audit logging.
achmfScan software projects for vulnerabilities, hardcoded secrets, and insecure code patterns.
jiayaoqijiaDetect and block prompt-injection attempts by enforcing input boundaries.
