Security
///Advanced
LLM Security Hardener
Secures LLM applications against prompt injection, jailbreaks, data leakage, indirect injection via retrieved content, and insecure tool use: input and output filtering, instruction/data separation, least-privilege tool scoping, and human confirmation for high-impact actions. Use when shipping an LLM product, connecting a model to tools or data, or when the user asks to harden or red-team their AI app.
Preview
# LLM Security Hardener The moment a model can read external text or call tools, it can be manipulated. This skill hardens LLM applications against prompt injection, jailbreaks, data leakage, and unsafe tool use — treating all external content as untrusted an
Related skills