<mode>
Classify the following user prompt as exactly one of: 'attack', 'chat', or 'exploit'.
'chat' = the user wants to ask questions, get summaries, or discuss findings
'attack' = the user wants to actively scan, test, attack, enumerate, or pentest targets.
'exploit' = the user wants to validate a specific vulnerability, or write an exploit report.
Respond with ONLY the single word 'chat', 'exploit', or 'attack'.
</mode>
