always-ask-next

To prompt an agent for the next action before ending a turn, you enforce a structured multiple-choice Next Action gate at the primary agent closeout moments. This prevents handoff ambiguity and ensures the agent confirms the user's intended next step before finalizing work.

To restrict scope control and stop subagents from prompting users, apply strict scope gating rules that ensure only the primary user-facing agent asks for Next Action. Background workers and subagents must return their integrated results directly without triggering user prompts.

Yes, agent orchestration supports MCP ask-user-questions flow for workflow gating by using it as the primary interface to present multiple-choice Next Action prompts. If the MCP flow is unavailable, the agent falls back to an equivalent in-message multiple-choice prompt.

The best way to enforce multi-select follow-through in agent closeout workflows is to apply control-flow guarantees within the Next Action prompt. This ensures structured execution of both single-choice selections and multi-select follow-through operations before the agent finalizes user-visible work.

Your multi-agent workflow stops too early because it lacks a Next Action gate to confirm the user's next step before finalizing the turn. Enforcing AUQ-first interaction rules at the primary agent closeout moment prevents these stopped too early outcomes.

Yes, you can use always-ask-next to manage background worker outputs by enforcing strict scope control. Background or subagent outputs are integrated directly into the primary agent closeout moments, allowing only the primary user-facing agent to trigger the Next Action prompt.