Skip to content

An AI-powered search engine with a generative UI

docs(prompts): enhance task completion verification in adaptive mode

Yoshiki Miurad93201dcfa98
diff --git a/lib/agents/prompts/search-mode-prompts.ts b/lib/agents/prompts/search-mode-prompts.ts
index f0d7b850..16a86ef9 100644
--- a/lib/agents/prompts/search-mode-prompts.ts
+++ b/lib/agents/prompts/search-mode-prompts.ts
@@ -219,8 +219,14 @@ TASK MANAGEMENT (todoWrite tool):
- Update status: pending → in_progress → completed
- This provides transparency and ensures nothing is missed
- **For complex queries (5+ steps), todoWrite becomes especially valuable for maintaining structure and ensuring comprehensive coverage**
- - For each task: set status to in_progress while working; when complete, set to completed via todoWrite. Keep statuses current.
- - Do not produce the final answer until you have called todoRead and verified completedCount equals totalCount. If not, continue executing or adjust the plan with todoWrite and verify again.
+
+**Task completion verification (CRITICAL):**
+- While working on a task: set its status to in_progress
+- When a task is complete: set it to completed via todoWrite after each meaningful step
+- Keep task statuses current throughout execution
+- **Before composing the final answer: you MUST call todoRead and verify completedCount equals totalCount**
+- If not all tasks are completed: continue executing remaining tasks or adjust the plan with todoWrite, then verify again with todoRead
+- Only proceed to write the final answer after confirming all tasks are completed
Example task patterns:
- "Research [topic] from multiple sources"