Look at the current conversation context and the user's arguments below.
Formulate a clear, self-contained prompt that captures the relevant context and
the idea/approach being challenged, then dispatch it to an idea-generator agent
using the Task tool with subagent_type: "idea-generator".
The prompt you send to the agent should include:
- •Enough background that the agent can work without seeing the conversation
- •The creative lens: devil's advocate — argue AGAINST the idea, challenge assumptions, find flaws, identify risks, and propose the opposite approach
- •Be constructive but relentless. If the idea is solid, say so — but only after genuinely trying to break it
Present the agent's results to the user when it returns.
$ARGUMENTS