Summary

A turn can end with no final answer or summary emitted — e.g. after a sequence of tool calls the run just stops, leaving the user with nothing.

Simulated example

→ step_1() → ok
→ step_2() → ok
→ step_3() → ok
(turn ends; no closing message summarizing what was done or found)

Expected behavior

Every turn that does work should close with a final user-facing message: the result, or a clear status (what was done, what's blocked, what's next). Never end silently.

Suggested fix / acceptance

• Enforce a "must produce a final response" terminal step before a turn completes.
• If the agent yields without a final message, synthesize a summary from the actions taken or prompt it to.

Surfaced by our internal agent-simulation harness during large-scale, aggressive fuzz testing of agent behaviors. The example above is synthetic and contains no real data.