AI article
Indirect prompt injection in a RAG pipeline: one attack, step by step, and what actually stopped it
Community description: Most write-ups about prompt injection show you the attacker typing something clever into the chat...
Dev.to | Sep 14, 2026 | sergio belmonte morales
Automated excerpt
This post walks through one such attack, recorded against a real model with the guardrails switched off, and then through what stopped it when they were switched back on. But it is a model-behaviour guardrail: it changes the prompt, and its effect only exists when a model is actually re-run against the new prompt. The same repository lists four attacks, and one of them says SKIP, not STOPPED: direct prompt injection, asking the assistant to print its own system prompt.
Selected automatically from source text; not independently written or fact-checked. Read the original for full context.