Prompt Injection: a working taxonomy
My own reference notes on the different flavors of prompt injection and where each shows up in real systems.
Prompt-Injection
Vulnerable Agent Lab: attacking a tool-using LLM agent
Building a deliberately vulnerable LangChain agent with file and email tools, then attacking it with indirect prompt injection.
Week 1: OWASP LLM Top 10 and Prompt Injection Testing on Llama 3 and Mistral
Testing three prompt injection techniques against two local LLMs (Llama 3 8B and Mistral 7B) — direct override, roleplay jailbreak, and encoding obfuscation — to see how they hold up.
Week 3: RAG Pipeline
Testing whether a RAG pipeline can be hijacked by an instruction hidden inside a retrieved document — and why a task-specific prompt template may accidentally provide some resistance.