adbc-drivers · eric-wang-1990 · Jul 31, 2026 · Jul 31, 2026 · Jul 31, 2026 · Jul 31, 2026
@@ -241,4 +241,9 @@ jobs:
           MODEL_ENDPOINT:   https://${{ secrets.DATABRICKS_HOST }}/serving-endpoints/databricks-claude-opus-4-8/invocations
           REPO_RULES_FILES: csharp/CLAUDE.md,rust/CLAUDE.md   # adbc keeps conventions per-language (no root CLAUDE.md)
           DATABRICKS_TOKEN: ${{ secrets.DATABRICKS_TOKEN }}
-        run: python -m databricks_bot_engine.engineer_bot.run --phase followup --bot .bot
+        # --system-prompt is REQUIRED for the followup phase (engine #89 removed the
+        # .bot/ prompt-name fallback). Point at the engineer-followup system prompt
+        # that already exists in .bot/prompts/ (mirrors databricks-jdbc's onboarding).
+        run: >-
+          python -m databricks_bot_engine.engineer_bot.run --phase followup --bot .bot
+          --system-prompt .bot/prompts/engineer-followup/system.md
@@ -55,12 +55,11 @@ concurrency:
 jobs:
   followup:
     # SECURITY: skip fork PRs — keep model/LLM secrets out of untrusted code's reach.
-    # Same `review-bot` label gate as reviewer-bot.yml — the follow-up only acts on
-    # PRs the reviewer was allowed to review.
+    # No label gate (matches reviewer-bot.yml and the other consumer repos): the
+    # follow-up acts on any open, non-fork PR, mirroring the initial reviewer.
     if: >-
       github.event.pull_request.head.repo.fork == false
       && github.event.pull_request.state == 'open'
-      && contains(github.event.pull_request.labels.*.name, 'review-bot')
     runs-on: [self-hosted, Linux, X64, peco-driver]
     timeout-minutes: 15
     steps:

@@ -32,12 +32,7 @@ name: Reviewer Bot
 
 on:
   pull_request:
-    # `labeled` is included so adding `review-bot` to an already-open PR triggers
-    # a review immediately, not just on the next push. (The `if:` gate below only
-    # honors `review-bot`; applying `engineer-bot` alone fires this event but the
-    # job is skipped. Engineer-bot fix PRs are reviewed because engineer-bot.yaml
-    # separately adds the `review-bot` label after publish.)
-    types: [opened, synchronize, reopened, ready_for_review, labeled]
+    types: [opened, synchronize, reopened, ready_for_review]
 
 permissions:
   contents: read
@@ -51,18 +46,14 @@ concurrency:
 
 jobs:
   review:
-    # Non-fork, non-draft, `review-bot`-labeled PRs only:
-    #  - fork PRs must never see the LLM/engine secrets; drafts aren't ready.
-    #  - label gate: review ONLY PRs carrying `review-bot`. `review-bot` guards the
-    #    reviewer; `engineer-bot` guards the engineer — strictly separate, never
-    #    mixed. A maintainer opts a PR in with `review-bot`; the bug-fix bot's own
-    #    fix PRs get `review-bot` applied by engineer-bot.yaml after publish, so
-    #    they're reviewed and the loop still closes. Without the label the reviewer
-    #    stays silent, so it doesn't comment on every PR in the repo.
+    # Non-fork, non-draft PRs — auto-review, matching the other consumer repos
+    # (databricks-sql-python/go/nodejs/kernel), which review every non-draft,
+    # non-fork PR without an opt-in label. fork == false keeps the LLM/engine
+    # secrets off untrusted code; drafts aren't ready. (Previously also required a
+    # `review-bot` label; dropped for fleet consistency.)
     if: >-
       github.event.pull_request.draft == false
       && github.event.pull_request.head.repo.fork == false
-      && contains(github.event.pull_request.labels.*.name, 'review-bot')
     # ACTIVATION BLOCKER: before taking this workflow out of DRAFT, switch this to
     # the protected `peco-driver` self-hosted runner (or adopt OIDC + a scoped,
     # short-lived model credential distinct from the CI DATABRICKS_TOKEN). As long