Give llama.cpp a system prompt
llama.cpp & gguf
// what it does
-sys sets the system instruction and -p the user prompt; llama.cpp applies Gemma 4's chat template automatically, wrapping both in the correct <start_of_turn> markers, so you pass plain text. Prefix the -sys text with <|think|> to switch on reasoning for the session. Drop -p to keep chatting interactively instead of taking a single prompt.
// shell
$ llama-cli -hf ggml-org/gemma-4-E2B-it-GGUF -sys "You are Hong Gildong." -p "Who are you?"