diff --git a/assets/skills/spawn.md b/assets/skills/spawn.md
index 0f58fa9..86912ea 100644
--- a/assets/skills/spawn.md
+++ b/assets/skills/spawn.md
@@ -9,17 +9,19 @@ The scope consists of all ready quests that are not claimed.
 Use the argument (if provided) to filter to specific quests/questlines.
 For a quest waiting on an outside condition, check whether it has cleared before recommending.
 
-Interactively prompt the user with one question per quest, never grouping quests into one question (a prompt may hold a few questions), with your recommendation:
+Present one quest at a time. Before its question, write a short paragraph explaining the problem, the quest's goal, and your recommendation, including any blockers or tradeoffs the user needs to decide.
+Ask one interactive question and wait for the answer before asking the next. Keep an asynchronous prompt active while waiting; do not end the turn while it awaits an answer.
+Offer these choices:
+
 - `/quest-start`: If the quest is well planned with no blockers.
 - `/quest-plan`: If the quest has significant design issues.
 - skip: If the quest should not be started yet.
 - deprioritize: If the quest should wait behind other work; `/quest-plan` moves it down.
 - `/quest-delete`: If the quest should be deleted.
 
-Include a brief summary of each quest.
 Consider ordering, suggesting to skip any quests that would result in conflicts.
 
-Spawn a background sub-agent for each `/quest-start` with its own isolated worktree, since agents cannot write to a worktree their parent created.
+Only after the user chooses `/quest-start` for that quest, spawn a background sub-agent with its own isolated worktree, since agents cannot write to a worktree their parent created.
 The agent switches it to the base `quest branch` prints (`git checkout -B <quest branch> origin/<base>`), creating the questline branch first if it is missing.
 
 Each agent reports back only when done or blocked.
@@ -30,9 +32,8 @@ When done, each agent opens a draft PR.
 Its report lists every open decision and suggested follow-up with a recommendation.
 Report each sub-agent's final status, staying silent on interim notifications, but do not monitor their PRs.
 
-As each agent reports, explain its result in a few lines, then interactively prompt the user.
+As each agent reports, explain its result in a few lines, then follow the prompt flow above.
 Include quest context, as many quests are concurrently in flight.
 
-Run any `/quest-plan` sessions in the foreground.
-Start any asynchronous tasks first before blocking on user prompt.
-Perform any research and monitoring in the background.
+Run selected `/quest-plan` sessions in the foreground.
+While awaiting the active prompt, continue independent research and already authorized background work.