feat(moments): a mount that keeps arriving where it does not apply proposes its own removal (milestone 458 step 7b, #4955)

The open-after-moment signal proposes a mount; nothing proposed taking one
off, so a wrong mount was noise at every occurrence until someone happened
to notice. rule_misfired(rule_id, moment, why, reached_by) records a report
against a MOUNTED pair, counted per distinct day (the MCP door carries no
session id) on a new rule_moment_judgments.misfire column (migration 0119,
backup v24). At three days the response carries a line asking the agent to
offer the operator the fix - reject takes the rule off, unmap_action stops
the action reaching the moment, confirm keeps the mount and stops the
asking - and Settings > Moments lists it as an unmount proposal with the
reasons and the actions that reached it. A re-mount clears the count.

Taught in moments.md, missed-retrieval.md and the reply hold's wording.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

src/scribe/mcp/server.py

@@ -218,6 +218,9 @@ _WRITE_TOOLS = frozenset({
     # Proposing moments for a rule writes a judgment row; judging one mounts
     # or unmounts the rule (step 7).
     "propose_rule_moments", "judge_rule_moments",
+    # A misfire report writes a counted row that can become an unmount
+    # proposal (step 7b).
+    "rule_misfired",
     # A reviewer's verdicts on logged menu lines (#4772) — rows carrying free
     # prose the agent authored, `rule_outcome`'s reason for being a write.
     "judge_menu",

src/scribe/mcp/tools/moments.py

@@ -147,10 +147,13 @@ async def propose_rule_moments(proposals: list[dict]) -> dict:
 async def rule_moment_proposals(rule_id: int = 0) -> dict:
     """The moment proposals waiting on the operator, grouped by rule.
 
-    Each rule carries what it is mounted on now and its proposals: the moment,
-    where the proposal came from (`pass` — read from the rule; `signal` — the
-    rule kept being opened just after that moment fired), the reason, and the
-    evidence counts. `rule_id` narrows to one rule.
+    Each rule carries what it is mounted on now and its proposals. Each
+    proposal is a `mount` or an `unmount`, with the moment, where it came
+    from (`pass` — read from the rule; `signal` — the rule kept being opened
+    just after that moment fired; `misfire` — sessions reported the mount
+    arriving where it did not apply, with their `reasons` and the actions
+    that `reached_by` the moment), the reason, and the evidence counts.
+    `rule_id` narrows to one rule.
     """
     return await judgments_svc.pending(current_user_id(), rule_id=rule_id or None)
 
@@ -160,14 +163,54 @@ async def judge_rule_moments(judgments: list[dict]) -> dict:
 
     Each item is `{"rule_id": N, "moment": "work.finish", "verdict":
     "confirm" | "reject", "note": "why"}`. Confirm mounts the rule on that
-    moment beside what it already has; reject records that it does not belong
-    there (and unmounts it if it was mounted), so neither the pass nor the
-    signal proposes the pair again. Moment `""` judges a "no moment fits"
+    moment beside what it already has — on an `unmount` proposal, it keeps
+    the mount and stops the misfire question; reject records that it does not
+    belong there (and unmounts it if it was mounted), so neither the pass nor
+    the signal proposes the pair again. Moment `""` judges a "no moment fits"
     answer. Put the operator's reason in `note`.
     """
     return await judgments_svc.judge(current_user_id(), judgments)
 
 
+async def rule_misfired(rule_id: int, moment: str, why: str,
+                        reached_by: str = "", project_id: int = 0) -> dict:
+    """Report that a rule MOUNTED on a moment arrived there and did not apply.
+
+    A mount delivers its rule every time the moment fires, whatever the work
+    is about — so a mount that is wrong is noise at every occurrence, and
+    nothing else notices. Call this when a line said a rule arrived *at* a
+    moment ("at work.verify, reached by `actions_run_read`"), you read it,
+    and it does not govern what you were doing there. It costs one call and
+    asks nothing of the operator; reports gather, counted once per day, and
+    once a pair has them on three distinct days the response carries a line
+    asking you to offer the operator the fix — take the rule off the moment,
+    or unmap the action when it is the action that is wrong here.
+
+    A rule that applied, even one you were already following, is not a
+    misfire. A rule that arrived by resemblance rather than a mount is
+    refused with the fix for that (its trigger).
+
+    Args:
+        rule_id: the rule the line named.
+        moment: the moment it arrived at, as the line names it.
+        why: what you were doing and why the rule did not bear on it. Required
+            — it is what tells the operator whether the rule or the action
+            is wrong.
+        reached_by: the action the line says reached the moment (`git push`,
+            `status=done`). Counted, so the operator can see which action
+            keeps bringing it.
+        project_id: the project you are working in (0 = none).
+
+    Returns `recorded`, `days` so far against the `bar`, and `context`: a line
+    to act on once the bar is crossed, else "". A mount the operator chose to
+    keep says so under `kept`.
+    """
+    return await judgments_svc.misfired(
+        current_user_id(), rule_id, moment, why=why, reached_by=reached_by,
+        project_id=project_id or None,
+    )
+
+
 def register(mcp) -> None:
     mcp.tool(name="list_moments")(list_moments)
     mcp.tool(name="map_action")(map_action)
@@ -176,3 +219,4 @@ def register(mcp) -> None:
     mcp.tool(name="propose_rule_moments")(propose_rule_moments)
     mcp.tool(name="rule_moment_proposals")(rule_moment_proposals)
     mcp.tool(name="judge_rule_moments")(judge_rule_moments)
+    mcp.tool(name="rule_misfired")(rule_misfired)

src/scribe/models/rule_moment_judgment.py

@@ -60,6 +60,11 @@ class RuleMomentJudgment(Base, CreatedAtMixin):
     note: Mapped[str | None] = mapped_column(Text, nullable=True)
     # The co-occurrence evidence (signal source), in lesson_rules' shape.
     evidence: Mapped[dict | None] = mapped_column(JSONB, nullable=True)
+    # The other direction (step 7b, migration 0119): sessions saying this
+    # MOUNTED rule arrived at the moment and did not apply. lesson_rules'
+    # evidence shape keyed per day, plus the reasons given and the actions
+    # that reached the moment; `kept_at` once the operator kept the mount.
+    misfire: Mapped[dict | None] = mapped_column(JSONB, nullable=True)
     judged_at: Mapped[datetime | None] = mapped_column(
         DateTime(timezone=True), nullable=True,
     )
@@ -76,6 +81,7 @@ class RuleMomentJudgment(Base, CreatedAtMixin):
             "source": self.source,
             "note": self.note or "",
             "evidence": self.evidence or {},
+            "misfire": self.misfire or {},
             "judged_at": iso(self.judged_at),
             "created_at": iso(self.created_at),
         }

src/scribe/services/backup.py

@@ -109,8 +109,11 @@ logger = logging.getLogger(__name__)
 # proposals waiting on a mount, and the rejections and "no moment fits"
 # answers that stop a pass or the open-after-moment signal proposing the same
 # pair again. Losing them puts every answered question back on the list.
+# v24 (2026-10) added rule_moment_judgments.misfire (milestone 458 step 7b):
+# the reports that a mounted rule arrived where it did not apply, and the
+# operator's "keep it" that stops them being proposed as an unmount again.
 # Bump when the serialized schema changes.
-BACKUP_VERSION = 23
+BACKUP_VERSION = 24
 
 # Every table this backup carries, by its REAL name. Paired with _NOT_INCLUDED
 # below, these two lists must together account for the entire schema — which is
@@ -772,6 +775,7 @@ def _rule_moment_judgment_rows(rows) -> list[dict]:
         {
             "rule_id": r.rule_id, "moment": r.moment, "state": r.state,
             "source": r.source, "note": r.note, "evidence": r.evidence,
+            "misfire": r.misfire,
             "judged_at": r.judged_at.isoformat() if r.judged_at else None,
             "created_at": r.created_at.isoformat() if r.created_at else None,
         }
@@ -1580,6 +1584,8 @@ def _build_rule_moment_judgment(row: dict, maps: _Maps) -> RuleMomentJudgment |
         source=row.get("source") or "pass",
         note=row.get("note") or None,
         evidence=row.get("evidence"),
+        # Archives before v24 carry no misfire reports.
+        misfire=row.get("misfire"),
         # Absent stays absent: a suggestion was never judged.
         judged_at=_dt_or_none(row.get("judged_at")),
         created_at=_dt(row.get("created_at")),

src/scribe/services/lesson_rules.py

@@ -486,17 +486,19 @@ def situation_key(arm: str, text: str) -> str:
     """A fingerprint of one situation, so a repeat counts once.
 
     `arm` is part of the key — "p" for a prompt, "w" for a file being written,
-    "s" for a session (the rule↔moment signal, milestone 458 step 7) —
+    "s" for a session (the rule↔moment signal, milestone 458 step 7), "d"
+    for a day (the misfire signal, step 7b, whose door carries no session) —
     because each names a different kind of situation and must not collide.
     On the prompt arm the text is the prompt: lowercased, split into word
     tokens of _MIN_TOKEN or more, de-duplicated and sorted, so the same ask
     re-sent with different spacing, punctuation, case or word order is one
     situation. On the write arm the text is the PATH, not the code: every
     edit to one file is one situation, however much the code differs between
-    them. The session arm keys on the session id as given. Empty when there
-    is nothing to key on.
+    them. The session and day arms key on the id or date as given — a date's
+    "10" and "05" are shorter than _MIN_TOKEN, so tokenising it would make
+    every day of a year one situation. Empty when there is nothing to key on.
     """
-    if arm in ("w", "s"):
+    if arm in ("w", "s", "d"):
         basis = (text or "").strip()
     else:
         tokens = sorted({t for t in re.findall(r"[a-z0-9]+", (text or "").lower())

src/scribe/services/moment_delivery.py

@@ -181,13 +181,22 @@ def reply_hold_reason(held: list[dict]) -> str:
         trigger = f"; it applies when {item['trigger']}" if item.get("trigger") else ""
         parts.append(f"“{item['title']}” ({why}{trigger}) — get_rule({item['rule_id']})")
     noun = "a standing rule" if len(held) == 1 else "standing rules"
+    # A mount that does not bear on this reply is a misfire worth one call:
+    # reports gather into an unmount proposal for the operator (step 7b).
+    mounted = [item for item in held if item.get("moment")]
+    misfire = (
+        " If one mounted on a moment has nothing to do with this reply, "
+        f"rule_misfired({mounted[0]['rule_id']}, \"{mounted[0]['moment']}\", why) "
+        "records that."
+        if mounted else ""
+    )
     return (
         f"Held for one read before this reply goes out: {noun} this session "
         f"has not opened: " + "; ".join(parts) + ". Read "
         + ("it" if len(held) == 1 else "them")
         + ", then send the reply — unchanged if it already does what the rule "
-        "asks, which is a judgement only you can make. This check runs once "
-        "per rule; the rewrite is never held."
+        "asks, which is a judgement only you can make." + misfire
+        + " This check runs once per rule; the rewrite is never held."
     )
 
 

src/scribe/services/rule_moment_judgments.py

@@ -19,6 +19,16 @@ Neither source mounts anything by itself. A suggestion that delivered its rule
 would manufacture the opens it counts, and the operator's corpus is theirs to
 mount.
 
+THE MISFIRE (step 7b) is the same machinery pointed the other way. A session
+that was handed a mounted rule at a moment where it did not apply says so
+(`misfired`), with why and the action that reached the moment. Reports count
+per distinct DAY — the MCP door is stateless and no session id reaches it, and
+a misfire seen on three days is a pattern where three in one afternoon may be
+one piece of work. At the bar the pair becomes an UNMOUNT proposal; the
+operator takes the rule off (reject), or unmaps the action, or keeps it
+(confirm), which stops the asking. A rule left unopened is never counted:
+most delivered rules go unopened, and the reply hold forces an open.
+
 EDITS ARE JUDGMENTS TOO. A person changing a rule's moments directly says
 which moments it belongs on, so `record_mount_change` (called from the one
 write path, `rulebooks.set_rule_moments`) confirms what was added and rejects
@@ -68,6 +78,17 @@ SIGNAL_WINDOW_SECONDS = 180
 # propose every rule onto both.
 SIGNAL_SKIP = frozenset({"work.run", "work.change"})
 
+# The reasons a misfire proposal carries, newest kept. Enough for the operator
+# to see whether the reports agree; few enough to read in the panel.
+MISFIRE_REASONS = 5
+# Distinct actions counted per pair. A moment is reached by a handful of
+# actions; past this the counts stop telling the operator which one to unmap.
+_MISFIRE_ACTIONS_CAP = 10
+_MISFIRE_WHY_CHARS = 400
+# On a mounted pair that has no judgment row — mounted before the rows were
+# kept (migration 0118) — when its first misfire gives it one.
+_UNRECORDED_MOUNT_NOTE = "mounted before judgments were recorded"
+
 
 def _clean_moment(name) -> str:
     """A catalog moment, normalised; "" stays "" (no moment fits)."""
@@ -279,22 +300,73 @@ async def pending(user_id: int, rule_id: int | None = None, limit: int = 200) ->
             "proposals": [],
         })
         entry["proposals"].append({
+            "proposal": "mount",
             "moment": judgment.moment,
             "source": judgment.source,
             "why": judgment.note or "",
             "evidence": lesson_rules.evidence_summary(judgment.evidence),
             "created_at": judgment.created_at.isoformat() if judgment.created_at else None,
         })
-    rules = list(grouped.values())
+    for judgment, rule, mounted in await _misfire_proposals(user_id, rule_id, limit):
+        entry = grouped.setdefault(rule.id, {
+            **_brief(rule),
+            "mounted": sorted(mounted, key=lambda n: (order.get(n, len(order)), n)),
+            "proposals": [],
+        })
+        entry["proposals"].append(_unmount_item(judgment))
+    rules = sorted(grouped.values(), key=lambda r: r["id"])
     return {"rules": rules, "total": sum(len(r["proposals"]) for r in rules)}
 
 
+def _misfire_due_clause():
+    """Proposed by the misfire signal and not yet kept by the operator."""
+    mf = RuleMomentJudgment.misfire
+    return mf["proposed_at"].astext.isnot(None), mf["kept_at"].astext.is_(None)
+
+
+async def _misfire_proposals(user_id: int, rule_id: int | None, limit: int):
+    """The unmount proposals: pairs past the misfire bar, unanswered, and
+    still mounted — a pair taken off since has had its answer."""
+    from scribe.services.rulebooks import _owned_rules_clause
+
+    async with async_session() as session:
+        stmt = (
+            select(RuleMomentJudgment, Rule)
+            .join(Rule, Rule.id == RuleMomentJudgment.rule_id)
+            .where(_owned_rules_clause(user_id), *_misfire_due_clause())
+        )
+        if rule_id:
+            stmt = stmt.where(Rule.id == int(rule_id))
+        rows = (await session.execute(
+            stmt.order_by(Rule.id, RuleMomentJudgment.moment).limit(limit)
+        )).all()
+        mounts = await _mounts(session, {r.id for _, r in rows})
+    return [(j, r, mounts.get(r.id, set())) for j, r in rows
+            if j.moment in mounts.get(r.id, set())]
+
+
+def _unmount_item(judgment: RuleMomentJudgment) -> dict:
+    mf = judgment.misfire or {}
+    reasons = list(mf.get("reasons") or [])
+    return {
+        "proposal": "unmount",
+        "moment": judgment.moment,
+        "source": "misfire",
+        "why": reasons[-1]["why"] if reasons else "",
+        "reasons": reasons,
+        "reached_by": dict(mf.get("reached_by") or {}),
+        "evidence": lesson_rules.evidence_summary(mf),
+        "created_at": mf.get("first_at"),
+    }
+
+
 async def judge(user_id: int, judgments: list[dict]) -> dict:
     """Confirm or reject proposals — or any (rule, moment) pair, proposed or not.
 
-    Confirm MOUNTS the rule on the moment (alongside what it already has);
-    reject records that it does not belong there and, if it was mounted,
-    unmounts it. Both go through `rulebooks.set_rule_moments`, the one write
+    Confirm MOUNTS the rule on the moment (alongside what it already has) —
+    on a pair already mounted, it KEEPS the mount and answers a misfire
+    proposal; reject records that it does not belong there and, if it was
+    mounted, unmounts it. Both go through `rulebooks.set_rule_moments`, the one write
     path, which records the judgment with this item's `note`. A moment of ""
     judges the "no moment fits" answer itself. A bad item is refused alone.
     """
@@ -382,11 +454,16 @@ async def record_mount_change(
 
     for moment in added:
         put(moment, CONFIRMED, _MOUNTED_NOTE)
+        # A fresh mount is a fresh judgment: misfires counted against an
+        # earlier mount of the pair are not evidence against this one.
+        rows[(rule_id, moment)].misfire = None
     for moment in removed:
         put(moment, REJECTED, _UNMOUNTED_NOTE)
     for moment in explicit:
         if verdict in (CONFIRMED, REJECTED):
             put(moment, verdict, "")
+        if verdict == CONFIRMED and moment in after:
+            _keep(rows[(rule_id, moment)], note, now)
     if after:
         none_row = rows.get((rule_id, NO_MOMENT))
         if none_row is not None and none_row.state == CONFIRMED:
@@ -394,6 +471,15 @@ async def record_mount_change(
             none_row.note = _OVERTURNED_NOTE
 
 
+def _keep(row: RuleMomentJudgment, note: str, now: datetime) -> None:
+    """The operator kept a mount the misfire signal proposed taking off. The
+    reports stay; the asking stops."""
+    mf = row.misfire
+    if not isinstance(mf, dict) or not mf.get("proposed_at") or mf.get("kept_at"):
+        return
+    row.misfire = {**mf, "kept_at": now.isoformat(), "kept_note": note or ""}
+
+
 # ── The signal ───────────────────────────────────────────────────────────────
 
 
@@ -525,3 +611,114 @@ async def opened_after(
             user_id, rule_id, moments, situation=session_id.strip(), project_id=project_id,
         )
     return {"moments": moments, "context": context}
+
+
+# ── The misfire: a mounted rule that arrived where it does not apply ─────────
+
+
+def misfire_situation(now: datetime) -> str:
+    """One situation per UTC day. The MCP door carries no session id; a day
+    is the coarser, honest unit, and the slower of the two to cross the bar."""
+    return lesson_rules.situation_key("d", now.date().isoformat())
+
+
+def add_misfire(misfire, key: str, project_id: int | None, now: datetime, *,
+                why: str, reached_by: str) -> dict:
+    """A NEW misfire dict with this report counted. Pure, like
+    `lesson_rules.add_evidence`, whose counting it reuses: the per-day
+    situations and the bar are the same as every other proposal's."""
+    ev = lesson_rules.add_evidence(misfire, key, project_id, now)
+    reasons = list(ev.get("reasons") or [])
+    reasons.append({"why": why[:_MISFIRE_WHY_CHARS], "reached_by": reached_by,
+                    "at": now.isoformat()})
+    ev["reasons"] = reasons[-MISFIRE_REASONS:]
+    actions = dict(ev.get("reached_by") or {})
+    if reached_by and (reached_by in actions or len(actions) < _MISFIRE_ACTIONS_CAP):
+        actions[reached_by] = int(actions.get(reached_by) or 0) + 1
+    ev["reached_by"] = actions
+    return ev
+
+
+def _unmount_line(rule: Rule, moment: str, misfire: dict) -> str:
+    days = len(misfire.get("situations") or [])
+    actions = ", ".join(f"`{a}` ×{n}" for a, n in sorted(
+        (misfire.get("reached_by") or {}).items(), key=lambda kv: -kv[1]))
+    kind = rule.kind or "rule"
+    return (
+        f"> {kind.capitalize()} #{rule.id} \"{rule.title}\" has been reported arriving at "
+        f"`{moment}` where it did not apply, on {days} distinct days"
+        + (f" (reached by {actions})" if actions else "")
+        + ". Offer the operator the fix in one line. If the rule does not belong "
+        f"at that moment, `judge_rule_moments([{{\"rule_id\": {rule.id}, \"moment\": "
+        f"\"{moment}\", \"verdict\": \"reject\", \"note\": \"why\"}}])` takes it off. "
+        "If it is the ACTION that is not that moment here, `unmap_action` for "
+        "it instead (`list_moments` shows the mapping). If they keep it, "
+        "`\"confirm\"` with their why stops the asking."
+    )
+
+
+async def misfired(
+    user_id: int, rule_id: int, moment: str, *, why: str,
+    reached_by: str = "", project_id: int | None = None,
+) -> dict:
+    """Record that a mounted rule arrived at `moment` where it did not apply.
+
+    Only a MOUNT can misfire: a rule that arrived by resemblance is a
+    retrieval question, answered by its trigger. `why` is required — a report
+    without its reason cannot tell the operator what to fix. Returns the
+    count so far and, once the pair crosses the bar, `context`: the line
+    asking the reader to offer the unmount. A kept mount still counts its
+    reports and asks nothing.
+    """
+    try:
+        name = moments_svc.require_moment((moment or "").strip())
+    except ValueError as exc:
+        return {"recorded": False, "error": str(exc)}
+    why = (why or "").strip() if isinstance(why, str) else ""
+    if not why:
+        return {"recorded": False, "error": (
+            "say why it did not apply here — the reason is what tells the "
+            "operator whether the rule or the action is wrong")}
+    rid = _int(rule_id)
+    reached_by = (reached_by or "").strip()[:200] if isinstance(reached_by, str) else ""
+    now = datetime.now(timezone.utc)
+    async with async_session() as session:
+        owned = await _owned_rules(session, user_id, [rid] if rid else [])
+        rule = owned.get(rid) if rid else None
+        if rule is None:
+            return {"recorded": False, "error": "not a rule you own"}
+        mounted = (await _mounts(session, [rid])).get(rid, set())
+        if name not in mounted:
+            return {"recorded": False, "error": (
+                f"rule {rid} is not mounted on {name}"
+                + (f" (it is on {', '.join(sorted(mounted))})" if mounted else "")
+                + ". A rule that arrived by its words rather than a mount is "
+                "fixed through its trigger — update_rule(when_to_apply=...).")}
+        row = (await _rows(session, [rid])).get((rid, name))
+        if row is None:
+            row = RuleMomentJudgment(rule_id=rid, moment=name, state=CONFIRMED,
+                                     source="edit", note=_UNRECORDED_MOUNT_NOTE,
+                                     judged_at=now)
+            session.add(row)
+        mf = add_misfire(row.misfire, misfire_situation(now), project_id, now,
+                         why=why, reached_by=reached_by)
+        kept = bool(mf.get("kept_at"))
+        due = not kept and lesson_rules.proposal_due(mf, now)
+        if due:
+            mf["proposed_at"] = now.isoformat()
+            mf["proposed_count"] = int(mf.get("proposed_count") or 0) + 1
+        row.misfire = mf
+        try:
+            await session.commit()
+        except IntegrityError:
+            await session.rollback()
+            return {"recorded": False, "error": "a report for this pair landed at the same moment; try again"}
+    out = {
+        "recorded": True, "rule_id": rid, "moment": name,
+        "days": len(mf.get("situations") or []),
+        "bar": lesson_rules.PROPOSE_SITUATIONS,
+        "context": _unmount_line(rule, name, mf) if due else "",
+    }
+    if kept:
+        out["kept"] = {"at": mf.get("kept_at"), "note": mf.get("kept_note") or ""}
+    return out