fix(task-review): accept org-native LAST_REVIEWED stamps, warn on bad ones

task-review-staleness.sh expected a bare 2026-07-09 and treated an org inactive timestamp ([2026-07-09 Thu], the form matching the CREATED:/CLOSED: cookies in the same drawer) as unparseable. The count branch folded that into the stale count, so a freshly-reviewed task reported as never-reviewed, and a full review pass never dropped the startup nudge. Both the bare and bracketed forms now normalize to the ISO date. A value that is neither warns loudly to stderr (file:line:value) and stays out of the count, since a data error shouldn't hide as "never reviewed." task-review.org documents the accepted format.

Tested red/green: 5 new bats cases (bracketed fresh and stale, list-mode sort by real date, malformed warns and is excluded). Full suite 273 ok.

1 files changed, 33 insertions, 7 deletions

diff --git a/.ai/scripts/task-review-staleness.sh b/.ai/scripts/task-review-staleness.sh
index ed43712..50e0257 100755
--- a/.ai/scripts/task-review-staleness.sh
+++ b/.ai/scripts/task-review-staleness.sh
@@ -19,9 +19,14 @@
 # deeper headings, and cookie-less headings are not review units.
 #
 # A task is stale (count mode), and sorts oldest (list mode), when its
-# :LAST_REVIEWED: property is missing or unparseable (NIL sorts first), or
-# when its age strictly exceeds the threshold (age > N days; age == N is
-# still fresh).
+# :LAST_REVIEWED: property is missing (NIL sorts first) or when its age
+# strictly exceeds the threshold (age > N days; age == N is still fresh).
+#
+# :LAST_REVIEWED: accepts a bare date (2026-07-09) or an org-native
+# timestamp ([2026-07-09 Thu] or <2026-07-09 Thu ...>); both normalize to
+# the ISO date. A value that is present but parses to neither is a data
+# error: the script warns loudly to stderr (file:line:value) and leaves it
+# out of the stale count rather than silently treating it as never-reviewed.
 
 set -euo pipefail
 
@@ -46,7 +51,7 @@ num="$2"
 # only the property drawer between a qualifying heading and the next heading
 # is scanned.
 extract_tasks() {
-  awk '
+  awk -v fname="$todo_file" '
     function flush() {
       if (in_task) printf "%s\t%s\t%s\n", hline, (have_lr ? lr : "NONE"), heading
     }
@@ -65,7 +70,21 @@ extract_tasks() {
       v = $0
       sub(/^[ \t]*:LAST_REVIEWED:[ \t]*/, "", v)
       sub(/[ \t]*$/, "", v)
-      lr = v; have_lr = 1
+      raw = v
+      # Accept org-native stamps: strip a leading [ or < (inactive/active
+      # timestamp bracket) and take the leading ISO date, so [2026-07-09 Thu]
+      # and 2026-07-09 both normalize to 2026-07-09.
+      sub(/^[[<]/, "", v)
+      if (match(v, /^[0-9][0-9][0-9][0-9]-[0-9][0-9]-[0-9][0-9]/)) {
+        lr = substr(v, RSTART, RLENGTH); have_lr = 1
+      } else {
+        # Present but unparseable — a data error, not "never reviewed". Warn
+        # loudly (file:line:value) instead of silently folding it into the
+        # stale count, where a re-review in the same bad format would never
+        # drop the number and nothing would explain why.
+        printf "task-review-staleness: %s:%d: unparseable :LAST_REVIEWED: %s (expected YYYY-MM-DD or [YYYY-MM-DD Day])\n", fname, NR, raw > "/dev/stderr"
+        lr = "INVALID"; have_lr = 1
+      }
       next
     }
     END { flush() }
@@ -98,9 +117,16 @@ while IFS=$'\t' read -r hline value heading; do
     continue
   fi
 
-  # Unparseable date → treat as NIL (stale).
+  # Malformed stamp: extract_tasks already warned. A data error is not
+  # "never reviewed", so it stays out of the stale count.
+  if [ "$value" = "INVALID" ]; then
+    continue
+  fi
+
+  # A shape-valid but impossible date (e.g. 2026-13-45) slips past the awk
+  # regex; warn and skip rather than silently counting it.
   if ! rev_epoch=$(date -d "$value" +%s 2>/dev/null); then
-    count=$((count + 1))
+    echo "task-review-staleness: $todo_file: unparseable :LAST_REVIEWED: $value" >&2
     continue
   fi