Trap-turn interventions showed a fresh-order hazard: Qwen3-235B often checked prerequisites and rebooked completed work unless a user-role directive explicitly said not to redo it, while many Qwen3.6-35B text-protocol errors were receipt-like booking lines written after prose that said no new work would be booked; this smaller-model mode nearly disappeared with real tool calls.

model: Qwen3-235B · variant: fresh work order for finished step · protocol: text

model: Qwen3-235B · variant: fresh work order for finished step · protocol: tools

model: Qwen3.6-35B · variant: fresh work order for finished step · protocol: text

model: Qwen3.6-35B · variant: fresh work order for finished step · protocol: tools

model: Qwen3-235B · variant: confirmation wording, same work order · protocol: text

model: Qwen3-235B · variant: confirmation wording, same work order · protocol: tools

model: Qwen3.6-35B · variant: confirmation wording, same work order · protocol: text

model: Qwen3.6-35B · variant: confirmation wording, same work order · protocol: tools

model: Qwen3-235B · variant: status-question wording, same work order · protocol: text

model: Qwen3-235B · variant: status-question wording, same work order · protocol: tools

model: Qwen3.6-35B · variant: status-question wording, same work order · protocol: text

model: Qwen3.6-35B · variant: status-question wording, same work order · protocol: tools