ID: fb670424-76f Priority: 95 Type: one_shot Status: open
CRITICAL: Debates failing because Theorist hits MAX_TOOL_ROUNDS without generating hypotheses. Recent debate SDA-2026-04-02-gap-aging-mouse-brain-v2-20260402 shows Theorist returned '[MAX TOOL ROUNDS REACHED]' causing subsequent personas to fail. Need to fix prompt engineering so Theorist generates hypotheses FIRST, uses tools only for specific verification. Root cause: current prompts encourage tool use too much, Claude prioritizes tool calls over text generation.
max_tool_rounds default from 15 to 3 in scidex_orchestrator.py:1109. Prompt already had CRITICAL WORKFLOW section strengthened in prior work. Only the tool round limit needed fixing.