fix(mind): record --apply auto-regenerates complete/index.md

draft/maintenance/mind/lifecycle_record_auto_index.md

@@ -0,0 +1,30 @@
+# lifecycle record auto-index
+
+Difficulty: trivial
+Autonomy: safe
+Priority: medium
+
+## Request (verbatim)
+
+Make PyAutoMind `scripts/lifecycle.py` `cmd_record` automatically regenerate and
+`git add` `complete/index.md` when `--apply` is passed, so the index can never
+drift stale after a ship.
+
+## Root cause
+
+`record --apply` writes the completion record and stages it, but leaves
+`complete/index.md` stale. The separate `index --apply` step is documented in the
+ship skills (`ship_library`/`ship_workspace`) but gets forgotten. The **Lifecycle
+Drift** workflow (`.github/workflows/lifecycle_drift.yml`) runs `index --check` on
+every direct push to `main` that touches `complete/**`; a forgotten regen fails
+the run and emails the maintainer on every ship.
+
+## Fix
+
+In `cmd_record`, under the `--apply` branch (after the record is written and the
+prompt folded), regenerate the index via the existing `_render_index` /
+`_existing_curated` helpers, write `complete/index.md`, and `git add` it — the
+same effect as `index --apply`, folded into `record` so the two steps can't drift
+apart. Behaviour-preserving except `record --apply` now also freshens the index.
+
+@PyAutoMind

scripts/lifecycle.py

@@ -198,6 +198,17 @@ def cmd_record(args) -> int:
                            capture_output=True, text=True)
         if r.returncode != 0:
             prompt.unlink(missing_ok=True)
+
+    # Freshen complete/index.md in the same step so a shipped record never
+    # leaves it stale — the Lifecycle Drift guard runs `index --check` on every
+    # push to main touching complete/**, and the separate `index --apply` step
+    # was easy to forget (failing runs + maintainer emails). Same effect as
+    # cmd_index --apply, folded in so the two can't drift apart.
+    INDEX_MD.write_text(_render_index(_existing_curated()))
+    subprocess.run(["git", "-C", str(ROOT), "add", str(INDEX_MD)],
+                   capture_output=True, text=True)
+    print(f"index: refreshed {INDEX_MD.relative_to(ROOT)} "
+          f"({len(_all_records())} records)")
     return 0