diff --git a/.claude/agents/onboarding.md b/.claude/agents/onboarding.md
index 1e6f74794..d030d723f 100644
--- a/.claude/agents/onboarding.md
+++ b/.claude/agents/onboarding.md
@@ -3,7 +3,7 @@ name: onboarding
 description: Orchestrates the full dataset onboarding workflow for Data Basis — from raw data → clean data → BigQuery → metadata in the backend. Spawn this agent when the user asks to onboard a dataset.
 ---
 
-# Data Basis Pipeline Agent
+# Data Basis Onboarding Agent
 
 This agent orchestrates the full dataset onboarding workflow for Data Basis: from raw data → clean data → BigQuery → metadata in the backend.
 
@@ -51,9 +51,14 @@ After step 8 succeeds, output a verification checklist and wait for explicit app
 ✓ Cloud tables: OK
 ✓ Verify at: https://development.basedosdados.org/dataset/<slug>
 
+Table order set: <list in order, or "default">
+OL order set per table: <summary, or "default">
+
 Reply "approved" to promote to prod, or describe what needs fixing.
 ```
 
+Note: ordering steps are handled inside `onboarding-metadata` (steps 7–8 of that command). If the human did not specify an order during step 8, you may ask here whether they want to adjust ordering before promoting to prod.
+
 Do not proceed to step 9 or step 10 without the user replying "approved" (or equivalent).
 
 ## Environment
diff --git a/.claude/commands/onboarding-metadata.md b/.claude/commands/onboarding-metadata.md
index 8baaf1231..6ee61dc10 100644
--- a/.claude/commands/onboarding-metadata.md
+++ b/.claude/commands/onboarding-metadata.md
@@ -68,6 +68,15 @@ For each entity in the table's observation level list (from architecture design
 
 Track returned OL IDs — needed for column updates.
 
+**Ordering checkpoint:** After all OLs for this table are created, ask the human:
+
+```text
+OLs created for <table_slug>: <entity_slug_1>, <entity_slug_2>, ...
+Should I set a specific display order for these observation levels? If yes, reply with the desired order (e.g. "year, municipality, financing_account"). Otherwise reply "default" to skip.
+```
+
+Wait for the reply before proceeding. If the human specifies an order, call `reorder_observation_levels` with the corresponding OL IDs in the requested order.
+
 ### 3. Upload columns from architecture table
 
 Call `upload_columns_from_sheet` with:
@@ -141,6 +150,17 @@ Call `create_update_table` again with the same `id` and `slug`, passing only:
 
 All other fields must be re-passed as well (the API requires them). This deferred call ensures the table record is fully persisted before the relationship is written.
 
+## Table ordering checkpoint
+
+After all tables are created/updated, ask the human:
+
+```text
+All tables registered for <dataset_slug>: <table_slug_1>, <table_slug_2>, ...
+Should I set a specific display order for these tables? If yes, reply with the desired order (one slug per line or comma-separated). Otherwise reply "default" to skip.
+```
+
+Wait for the reply. If the human specifies an order, call `reorder_tables` with the requested slug order.
+
 ## Summary output
 
 After processing all tables, output: