PromptScan

This gives your AI agent a way to check for prompt injection attacks before feeding user input to an LLM. It connects to the PromptScan detection API and lets you scan arbitrary text strings for malicious prompting patterns. You'd reach for this when building agents that handle untrusted input, like customer support bots, content moderators, or any system where users might try to manipulate your prompts. The server uses streamable HTTP transport and hits the promptscan.dev endpoints to run the actual detection. It's a straightforward security layer: send text in, get a risk assessment back, decide whether to proceed or reject the input.