intake: put the backlog in the Recent feed

agents/conductors/intake/AGENTS.md

@@ -90,7 +90,11 @@ schema — light structure over free-form prose.
 
 **Recent** is the one section laid out by *date* rather than by state: the 50
 newest events on the **work in hand** — issued, parked, filed — merged across
-the live buckets and sitting between the Backlog and the Epics. It *holds* 50
+every live bucket (the `draft/` backlog included, which is most of them:
+150 prompts against a handful of registry rows) and sitting between the Backlog
+and the Epics. Epic members stay out, as they do in every pick list on the
+page — they are worked in order through their epic, and a Recent row hands out
+a standalone `/start_dev`. It *holds* 50
 and *shows* 10 (`RECENT_MAX` / `RECENT_PAGE`): the table is a glance, not a
 log, so the rest is one tap away — a `…` button on the Pages twin, and nested
 `<details>` on the markdown page, which GitHub renders where it strips the

agents/conductors/intake/_intake.py

@@ -337,14 +337,15 @@ def parse_header(text: str) -> dict:
 
     Only scans the top of the file so a stray "Status:" deep in prose does not
     fire; first occurrence of each field wins. No YAML — the blessed convention.
-    `Epic:`/`Phase:` are optional epic-membership fields (dashboard grouping)
-    and `Issued:` is the prompt's own copy of its registry date; none are in
+    `Epic:`/`Phase:` are optional epic-membership fields (dashboard grouping);
+    `Filed:`/`Issued:` are the prompt's own date, keyed by the state it was in
+    when that happened (PyAutoMind REFERENCE.md "Task dates"). None are in
     HEADER_FIELDS, so their absence is never header hygiene.
     """
     fields = {}
     for line in text.splitlines()[:30]:
         m = re.match(r"(Type|Target|Difficulty|Autonomy|Priority|Status|"
-                     r"Issued|Epic|Phase):\s*(\S.*)",
+                     r"Issued|Filed|Epic|Phase):\s*(\S.*)",
                      line.strip())
         if m:
             fields.setdefault(m.group(1).lower(), m.group(2).strip())
@@ -377,6 +378,18 @@ def _prefix_match(path: str, prefix: str) -> bool:
              "found", "completed", "shipped")
 
 
+def _header_date(header: dict) -> str:
+    """A prompt's own date from its `Issued:` / `Filed:` header, else ''.
+
+    `Issued:` wins when a prompt carries both — it is the later, more specific
+    event, and an issued prompt keeps the `Filed:` it had as a draft."""
+    for key in ("issued", "filed"):
+        m = _ISO_DATE.search(header.get(key) or "")
+        if m:
+            return m.group(1)
+    return ""
+
+
 def _entry_date(fields: dict) -> tuple:
     """(date, event) for a registry entry, or ('', '') when it carries none."""
     for key in DATE_KEYS:
@@ -537,6 +550,16 @@ def recent_events(c: dict, limit: int = RECENT_MAX) -> list:
             events.append({"date": r["date"], "event": r.get("event") or "issued",
                            "title": r["title"], "path": r["path"],
                            "payload": f"/start_dev {r['path']}"})
+    # The backlog is the LARGEST pool of work the Mind holds — 150 prompts
+    # against a handful of live rows — so a feed that skipped it could see
+    # almost none of what has been happening. Epic members stay out, as they do
+    # in every pick list on the page: they are worked in order through their
+    # epic, and a Recent row hands out a standalone `/start_dev`.
+    for r in c.get("records") or []:
+        if r.get("date") and not r.get("epic"):
+            events.append({"date": r["date"], "event": "filed",
+                           "title": r["title"], "path": r["path"],
+                           "payload": f"/start_dev {r['path']}"})
     for key, verb in (("parked", "resume"), ("planned", "start")):
         for e in c.get(key) or []:
             if e.get("date"):
@@ -593,6 +616,9 @@ def census(mind: Path) -> dict:
                 "status": header.get("status", "-"),
                 "epic": header.get("epic", ""),
                 "phase": phase,
+                # `Filed:` normally; `Issued:` only on a prompt that has been
+                # issued and moved back, which is still the later event.
+                "date": _header_date(header),
                 "header": header,
                 "missing": missing,
             })
@@ -639,8 +665,7 @@ def _count(key):
         # claims it) dated rather than dropping it out of the recent feed.
         date, event = row.get("date", ""), row.get("event", "")
         if not date:
-            m = _ISO_DATE.search(header.get("issued", ""))
-            date, event = (m.group(1), "issued") if m else ("", "")
+            date, event = _header_date(header), "issued"
         in_flight.append({
             "path": rel,
             "title": _title(text),

tests/test_intake_dashboard.py

@@ -675,3 +675,49 @@ def test_the_html_blurb_renders_its_code_spans(tmp_path):
     html = _intake.render_dashboard_html(_intake.census(_many(tmp_path, 80)))
     assert "<code>complete/index.md</code>" in html
     assert "`complete/index.md`" not in html
+
+
+# --------------------------------------------------------------------------- #
+# recent: the backlog is most of the work
+# --------------------------------------------------------------------------- #
+def test_a_dated_draft_is_in_the_feed(tmp_path):
+    """The backlog is the largest pool of work the Mind holds, so a feed that
+    skipped it saw almost none of what has been happening."""
+    mind = _mind(tmp_path, drafts={
+        "feature/widgets/sprocket.md":
+            _prompt("Sprocket work").replace("Status: formalised",
+                                             "Status: formalised\nFiled: 2026-08-20")})
+    rows = _intake.census(mind)["recent"]
+    assert [(r["date"], r["event"], r["title"]) for r in rows] == [
+        ("2026-08-20", "filed", "Sprocket work")]
+    assert rows[0]["payload"] == "/start_dev draft/feature/widgets/sprocket.md"
+
+
+def test_an_undated_draft_stays_out(tmp_path):
+    mind = _mind(tmp_path, drafts={"feature/widgets/sprocket.md": _prompt("S")})
+    assert _intake.census(mind)["recent"] == []
+
+
+def test_an_epic_member_is_not_offered_standalone_in_the_feed(tmp_path):
+    """Members are worked in order through their epic — every other pick list
+    on the page excludes them, and a Recent row hands out a `/start_dev`."""
+    member = _epic_prompt_body("Phase one", "jax-profiling", phase=1).replace(
+        "Status: formalised", "Status: formalised\nFiled: 2026-08-20")
+    mind = _mind(tmp_path, registries={"epics.md": _EPICS},
+                 drafts={"feature/widgets/phase_one.md": member,
+                         "feature/widgets/loose.md":
+                             _prompt("Loose end").replace(
+                                 "Status: formalised",
+                                 "Status: formalised\nFiled: 2026-08-21")})
+    assert [r["title"] for r in _intake.census(mind)["recent"]] == ["Loose end"]
+
+
+def test_issued_beats_filed_on_a_prompt_carrying_both(tmp_path):
+    """An issued prompt keeps the `Filed:` it had as a draft; the later, more
+    specific event is the one the feed reports."""
+    body = _prompt("Sprocket").replace(
+        "Status: formalised",
+        "Status: formalised\nFiled: 2026-07-01\nIssued: 2026-08-19")
+    rows = _intake.census(
+        _mind(tmp_path, active={"sprocket.md": body}))["recent"]
+    assert [(r["date"], r["event"]) for r in rows] == [("2026-08-19", "issued")]