04.7.9 · Skill in AI-native engineering
Guardrails
Checks that keep AI output safe, on topic and within policy.
Guardrails validate inputs and outputs, block unsafe content and limit what an agent can do. They range from simple rules to classifier models. They matter most where AI output reaches users or takes actions directly.
Resources
3- docsStart hereOWASP top 10 for LLM applications ↗OWASP GenAI Security Project · index
The main security risks of LLM features, with prompt injection at the top.
- docsStart herePrompt injection ↗OWASP GenAI Security Project · 2026 · 5 min
Why prompt injection cannot be patched away, and the controls that limit the damage.
- articleGo deeperPrompt injection (series) ↗Simon Willison · series
The clearest ongoing writing on prompt injection, from the person who named it.
Tools
none yetTools for this topic are being researched.