From f5fa7ec70f02b3baea98089e32bccbe81e7d6d21 Mon Sep 17 00:00:00 2001
From: jsahn <jsahn@eroomai.com>
Date: Sat, 12 Sep 2026 23:52:31 +0900
Subject: [PATCH] Add a Monday-morning weekly recap briefing
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

No US session closes into Monday KST, so Mondays (default 07:00, cron
in the daemon with restart catch-up) now publish a weekly review built
entirely from stored data: the quotes table collapsed into week-over-
week changes and the week's daily briefing headlines as the narrative.
The recap is a BriefingResult kind="weekly" reusing MarketPayload, so
the store, verification pass, all delivery channels (Blogger title
"미국 주간 리뷰 <monday>") and the dashboard (주간 리뷰 badge) work
unchanged. Manual runs via `run-once --weekly [--date <monday>]`.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---
 .env.example                  |   3 +
 README.md                     |   3 +
 dashboard/src/client/main.ts  |  11 +-
 dashboard/src/server.ts       |   1 +
 dashboard/src/shared/types.ts |   3 +
 dashboard/web/styles.css      |   1 +
 tests/_orchestration_stubs.py |  11 +
 tests/test_blogger.py         |   8 +
 tests/test_llm.py             |  56 ++++-
 tests/test_weekly.py          | 380 ++++++++++++++++++++++++++++++++++
 us2kr/cli.py                  |   9 +-
 us2kr/config.py               |   3 +
 us2kr/daemon.py               |  63 ++++++
 us2kr/deliver/blogger.py      |   5 +-
 us2kr/llm/base.py             |  18 +-
 us2kr/llm/minimax.py          |  33 ++-
 us2kr/llm/prompts.py          |  69 +++++-
 us2kr/models.py               |   3 +
 us2kr/render.py               |  39 ++++
 us2kr/store.py                |  22 ++
 us2kr/summarize.py            |  45 +++-
 us2kr/weekly.py               | 242 ++++++++++++++++++++++
 22 files changed, 995 insertions(+), 33 deletions(-)
 create mode 100644 tests/test_weekly.py
 create mode 100644 us2kr/weekly.py

diff --git a/.env.example b/.env.example
index 500632a..7dce34e 100644
--- a/.env.example
+++ b/.env.example
@@ -32,4 +32,7 @@ US2KR_SKIP_WHEN_KRX_CLOSED=false
 # exchange_calendars에 없는 한국 임시휴장일 (CSV, 예: 2026-06-03)
 US2KR_KRX_EXTRA_HOLIDAYS=
 US2KR_PREMARKET_UPDATE_ENABLED=false
+# 월요일 아침 주간 리뷰 (지난주 시세·헤드라인 요약)
+US2KR_WEEKLY_ENABLED=true
+US2KR_WEEKLY_TIME_KST=07:00
 US2KR_LOG_LEVEL=INFO
diff --git a/README.md b/README.md
index 6fc8cec..5df0bad 100644
--- a/README.md
+++ b/README.md
@@ -17,6 +17,9 @@
 | **브리핑 발송** (마감 +75분, 기본) | 17:15 | **06:15 / 07:15** |
 | 한국 개장 | — | 09:00 |
 
+- **월요일 아침(기본 07:00 KST)에는 주간 리뷰 발행** — 월요일 새벽엔 마감되는 미국 세션이 없으므로,
+  저장된 지난주 시세(주간 변동)와 일일 브리핑 헤드라인으로 주간 요약을 만들어 같은 채널로 발행
+  (`US2KR_WEEKLY_ENABLED` / `US2KR_WEEKLY_TIME_KST`, 수동: `run-once --weekly [--date 월요일]`)
 - 시각은 하드코딩이 아니라 `exchange_calendars`(XNYS/XKRX)로 계산 → 미국 휴장·조기폐장·서머타임 자동 대응
 - 미국 휴장일은 스킵, 한국 휴장일은 리포트에 "오늘 한국 휴장" 표기. `US2KR_SKIP_WHEN_KRX_CLOSED=true`면
   한국 휴장 연휴 중 **다음 미국 세션이 먼저 마감되어 무의미해지는 브리핑만** 스킵 — 금요일 세션(토요일 도착)은
diff --git a/dashboard/src/client/main.ts b/dashboard/src/client/main.ts
index e357fd8..6454723 100644
--- a/dashboard/src/client/main.ts
+++ b/dashboard/src/client/main.ts
@@ -166,7 +166,11 @@ function renderList(
     button.setAttribute("aria-current", String(s.krx_date === selected));
     button.append(el("span", "briefing-date", `${s.krx_date} (KRX)`));
     const line = el("span", "briefing-line");
-    line.append(`세션 ${s.session_date}`, badgeFor(s));
+    if (s.kind === "weekly") {
+      line.append(el("span", "badge badge-week", "주간 리뷰"), `~${s.session_date}`, badgeFor(s));
+    } else {
+      line.append(`세션 ${s.session_date}`, badgeFor(s));
+    }
     if (s.data_gaps.length > 0) line.append(el("span", "badge badge-warn", `결측 ${s.data_gaps.length}`));
     button.append(line);
     if (s.headline) button.append(el("span", "briefing-line", s.headline));
@@ -179,8 +183,11 @@ function renderList(
 function metaChips(result: BriefingResult): HTMLElement[] {
   const p = result.payload;
   const close = new Date(p.us_close_time);
+  const weekly = result.kind === "weekly";
   const chips: HTMLElement[] = [
-    el("span", undefined, `미국 세션 ${p.session_date} → KRX ${p.krx_date}`),
+    weekly
+      ? el("span", "badge badge-week", "주간 리뷰")
+      : el("span", undefined, `미국 세션 ${p.session_date} → KRX ${p.krx_date}`),
     el(
       "span",
       undefined,
diff --git a/dashboard/src/server.ts b/dashboard/src/server.ts
index 15c34f4..0c7b948 100644
--- a/dashboard/src/server.ts
+++ b/dashboard/src/server.ts
@@ -67,6 +67,7 @@ async function listDates(): Promise<string[]> {
 function summarize(krxDate: string, r: BriefingResult): BriefingSummary {
   return {
     krx_date: krxDate,
+    kind: r.kind ?? "briefing",
     session_date: r.payload.session_date,
     generated_at: r.payload.generated_at,
     krx_open_today: r.payload.krx_open_today,
diff --git a/dashboard/src/shared/types.ts b/dashboard/src/shared/types.ts
index 551dc09..7ba156d 100644
--- a/dashboard/src/shared/types.ts
+++ b/dashboard/src/shared/types.ts
@@ -93,12 +93,15 @@ export interface BriefingResult {
   error: string | null;
   /** Absent/null: verification not run (disabled, degraded, verifier failed, or pre-feature archive). */
   verification?: Verification | null;
+  /** "weekly": Monday recap (quotes carry week-over-week changes). Absent on older archives. */
+  kind?: "briefing" | "weekly";
 }
 
 // ---------------------------------------------------------------- API shapes
 
 export interface BriefingSummary {
   krx_date: string;
+  kind: "briefing" | "weekly";
   session_date: string;
   generated_at: string;
   krx_open_today: boolean;
diff --git a/dashboard/web/styles.css b/dashboard/web/styles.css
index 3c3d9f2..f824b62 100644
--- a/dashboard/web/styles.css
+++ b/dashboard/web/styles.css
@@ -205,6 +205,7 @@ button[aria-pressed="true"] { background: var(--grid); }
 .badge-ok { background: var(--badge-ok-bg); color: var(--badge-ok-ink); }
 .badge-warn { background: var(--badge-warn-bg); color: var(--badge-warn-ink); }
 .badge-error { background: var(--badge-err-bg); color: var(--badge-err-ink); }
+.badge-week { border: 1px solid var(--series-1); color: var(--series-1); background: transparent; }
 
 /* ------------------------------------------------------------------ summary card */
 .summary-card {
diff --git a/tests/_orchestration_stubs.py b/tests/_orchestration_stubs.py
index 2baa2a2..de8dc2d 100644
--- a/tests/_orchestration_stubs.py
+++ b/tests/_orchestration_stubs.py
@@ -152,6 +152,9 @@ class FakeStore:
         self.fail_history = fail_history
         self.closed = False
         self._next_id = 1
+        # weekly-recap sources: dict rows mimicking sqlite3.Row's __getitem__ access
+        self.quote_rows: list[dict] = []
+        self.report_rows: list[dict] = []
 
     def save_quotes(self, session_date: date, quotes: list) -> None:
         self.quotes[session_date] = list(quotes)
@@ -177,6 +180,14 @@ class FakeStore:
     def save_raw(self, session_date: date, name: str, payload_json: str) -> None:
         self.raw[(session_date, name)] = payload_json
 
+    def load_quotes_between(self, start: date, end: date) -> list[dict]:
+        rows = [r for r in self.quote_rows if start.isoformat() <= r["session_date"] <= end.isoformat()]
+        return sorted(rows, key=lambda r: (r["symbol"], r["session_date"]))
+
+    def reports_between(self, start: date, end: date, kind: str = "briefing") -> list[dict]:
+        rows = [r for r in self.report_rows if start.isoformat() <= r["krx_date"] <= end.isoformat()]
+        return sorted(rows, key=lambda r: r["krx_date"])
+
     def close(self) -> None:
         self.closed = True
 
diff --git a/tests/test_blogger.py b/tests/test_blogger.py
index 3df4dfe..4866d8c 100644
--- a/tests/test_blogger.py
+++ b/tests/test_blogger.py
@@ -255,3 +255,11 @@ def test_build_deliverers_includes_blogger_when_configured(tmp_path):
     blogger = deliverers[-1]
     assert blogger.labels == ["미국증시", "시황"]  # type: ignore[attr-defined]
     assert blogger.blog_id == "1"  # type: ignore[attr-defined]
+
+
+def test_post_title_weekly_kind():
+    result = make_result()
+    result = result.model_copy(update={"kind": "weekly"})
+    assert post_title(result) == "미국 주간 리뷰 2026-09-01 — 9월 첫 거래일 하락"
+    degraded = make_result(headline=None).model_copy(update={"kind": "weekly"})
+    assert post_title(degraded).startswith("미국 주간 리뷰 2026-09-01 (미 ")
diff --git a/tests/test_llm.py b/tests/test_llm.py
index d3f541f..e1e5028 100644
--- a/tests/test_llm.py
+++ b/tests/test_llm.py
@@ -645,12 +645,16 @@ class GlueProvider:
     def summarize(self, payload: MarketPayload) -> tuple[Report, str]:
         raise AssertionError("summarize must not be called")
 
-    def verify(self, payload: MarketPayload, report: Report) -> tuple[Verification, str]:
+    def verify(
+        self, payload: MarketPayload, report: Report, data_message: str | None = None
+    ) -> tuple[Verification, str]:
         self.verify_calls += 1
+        self.last_data_message = data_message
         return self._verdicts.pop(0), "stub"
 
-    def revise(self, payload, report, issues) -> tuple[Report, str]:
+    def revise(self, payload, report, issues, *, system_prompt=None, data_message=None) -> tuple[Report, str]:
         self.revise_calls += 1
+        self.last_system_prompt = system_prompt
         if self._revise_fails:
             raise LLMError("revise down")
         assert self._revised is not None
@@ -696,3 +700,51 @@ def test_verify_and_revise_revision_failure_keeps_original() -> None:
     out, verification = verify_and_revise(make_payload(), report, provider)
     assert out is report and verification is verdict
     assert provider.verify_calls == 1 and provider.revise_calls == 1
+
+
+# ---------------------------------------------------------------------- weekly summarize
+def test_summarize_weekly_uses_weekly_prompt_and_schema() -> None:
+    from datetime import date as _date
+
+    from us2kr.llm.prompts import WEEKLY_SYSTEM_PROMPT
+
+    provider, client = make_provider([make_completion(json.dumps(GOOD_REPORT, ensure_ascii=False))])
+    report, model = provider.summarize_weekly(make_payload(), [_date(2026, 3, 9), _date(2026, 3, 10)])
+
+    assert report.headline and model == "MiniMax-M3"
+    call = client.calls[0]
+    assert call["messages"][0]["content"] == WEEKLY_SYSTEM_PROMPT
+    assert "## 주간 시세" in call["messages"][1]["content"]
+    assert call["response_format"]["json_schema"]["name"] == "report"
+
+
+def test_summarize_weekly_failure_tags_task() -> None:
+    provider, _ = make_provider([make_status_error(500)] * 3)
+    from datetime import date as _date
+
+    with pytest.raises(LLMError, match=r"\(summarize-weekly\)"):
+        provider.summarize_weekly(make_payload(), [_date(2026, 3, 9)])
+
+
+def test_verify_accepts_explicit_data_message() -> None:
+    provider, client = make_provider([make_completion(json.dumps({"ok": True, "issues": []}))])
+    provider.verify(make_payload(), Report.model_validate(GOOD_REPORT), "## 주간 데이터\n- 지표")
+    user = client.calls[0]["messages"][1]["content"]
+    assert user.startswith("## 주간 데이터")
+    assert "## 검증 대상 브리핑 초안 (JSON)" in user
+
+
+def test_revise_accepts_weekly_persona_and_data() -> None:
+    from us2kr.llm.prompts import WEEKLY_SYSTEM_PROMPT
+
+    provider, client = make_provider([make_completion(json.dumps(GOOD_REPORT, ensure_ascii=False))])
+    provider.revise(
+        make_payload(),
+        Report.model_validate(GOOD_REPORT),
+        [VerificationIssue(claim="c", problem="p", severity="major")],
+        system_prompt=WEEKLY_SYSTEM_PROMPT,
+        data_message="## 주간 데이터",
+    )
+    messages = client.calls[0]["messages"]
+    assert messages[0]["content"] == WEEKLY_SYSTEM_PROMPT
+    assert messages[1]["content"] == "## 주간 데이터"
diff --git a/tests/test_weekly.py b/tests/test_weekly.py
new file mode 100644
index 0000000..d133639
--- /dev/null
+++ b/tests/test_weekly.py
@@ -0,0 +1,380 @@
+"""Weekly recap tests: collaborators are monkeypatched on us2kr.weekly; no network, no real DB."""
+
+from __future__ import annotations
+
+import json
+from datetime import date, datetime, timedelta
+from zoneinfo import ZoneInfo
+
+import pytest
+from _orchestration_stubs import FakeClock, FakeStore
+
+from us2kr import weekly
+from us2kr.config import Settings
+from us2kr.llm.base import LLMError
+from us2kr.llm.prompts import WEEKLY_SYSTEM_PROMPT, build_weekly_user_message
+from us2kr.models import Report, Verification
+from us2kr.pipeline import SKIP_ALREADY_SENT
+from us2kr.render import render_weekly_markdown
+from us2kr.weekly import build_weekly_payload, most_recent_monday, run_weekly, week_sessions_for
+
+KST = ZoneInfo("Asia/Seoul")
+MONDAY = date(2026, 9, 14)
+WEEK = [date(2026, 9, 7) + timedelta(days=i) for i in range(5)]  # Mon 9/7 .. Fri 9/11
+NOW = datetime(2026, 9, 14, 7, 5, tzinfo=KST)
+
+
+def make_settings(tmp_path, **overrides) -> Settings:
+    values = {
+        "US2KR_DATA_DIR": str(tmp_path / "data"),
+        "US2KR_REPORTS_DIR": str(tmp_path / "reports"),
+        "US2KR_LOG_DIR": str(tmp_path / "logs"),
+        "MINIMAX_API_KEY": "sk-test-not-a-real-key",
+    }
+    values.update(overrides)
+    return Settings(_env_file=None, **values)
+
+
+def quote_row(session: date, symbol: str, close: float, prev_close: float, **extra) -> dict:
+    row = {
+        "session_date": session.isoformat(),
+        "symbol": symbol,
+        "close": close,
+        "prev_close": prev_close,
+        "change_pct": None,
+        "change_abs": None,
+        "day_high": extra.get("day_high"),
+        "day_low": extra.get("day_low"),
+        "volume": None,
+    }
+    return row
+
+
+def report_row(krx_date: date, session: date, headline: str | None) -> dict:
+    result = {"payload": {"session_date": session.isoformat()}}
+    result["report"] = {"headline": headline} if headline is not None else None
+    return {
+        "krx_date": krx_date.isoformat(),
+        "session_date": session.isoformat(),
+        "result_json": json.dumps(result, ensure_ascii=False),
+    }
+
+
+def sample_report() -> Report:
+    return Report(
+        headline="반도체 랠리로 주간 상승",
+        market_overview="한 주간 지수가 올랐다.",
+        key_drivers=["엔비디아 실적"],
+        rates_fx_commodities="금리는 주간 상승.",
+        korea_implications=["국내 반도체 우호적"],
+        watchlist_today=["삼성전자"],
+        risk_flags=["금리 재상승"],
+    )
+
+
+def populated_store() -> FakeStore:
+    store = FakeStore()
+    # ^GSPC: Mon prev_close 100 -> Fri close 110 = +10% weekly
+    for i, session in enumerate(WEEK):
+        store.quote_rows.append(
+            quote_row(session, "^GSPC", 102.0 + 2 * i, 100.0 + 2 * i, day_high=103.0 + 2 * i, day_low=99.0)
+        )
+    # daily headlines Tue..Sat (krx dates for Mon..Fri sessions)
+    for session in WEEK:
+        store.report_rows.append(report_row(session + timedelta(days=1), session, f"{session.day}일 헤드라인"))
+    return store
+
+
+# ------------------------------------------------------------------- date helpers
+def test_most_recent_monday():
+    assert most_recent_monday(MONDAY) == MONDAY
+    assert most_recent_monday(date(2026, 9, 18)) == MONDAY  # Friday -> that week's Monday
+    assert most_recent_monday(date(2026, 9, 13)) == date(2026, 9, 7)  # Sunday -> previous Monday
+
+
+def test_week_sessions_excludes_holidays():
+    clock = FakeClock()
+    assert week_sessions_for(MONDAY, clock) == WEEK
+
+    class HolidayClock(FakeClock):
+        def is_us_session(self, d: date) -> bool:
+            return super().is_us_session(d) and d != date(2026, 9, 7)  # Labor-day-like Monday
+
+    assert week_sessions_for(MONDAY, HolidayClock()) == WEEK[1:]
+
+
+# ------------------------------------------------------------------- payload builder
+def test_build_weekly_payload_aggregates_week(tmp_path):
+    store = populated_store()
+    payload, sessions = build_weekly_payload(MONDAY, make_settings(tmp_path), FakeClock(), store)
+
+    assert sessions == WEEK
+    assert payload.krx_date == MONDAY and payload.session_date == WEEK[-1]
+    assert payload.krx_open_today is True
+    (q,) = payload.quotes
+    assert q.symbol == "^GSPC" and q.name == "S&P 500" and q.group == "indices"
+    assert q.prev_close == 100.0 and q.close == 110.0
+    assert q.change_pct == pytest.approx(10.0)
+    assert q.change_abs == pytest.approx(10.0)
+    assert q.day_high == 111.0 and q.day_low == 99.0
+    assert payload.history_notes == [f"미 {s.isoformat()}({w}): {s.day}일 헤드라인" for s, w in zip(WEEK, "월화수목금")]
+    assert payload.data_gaps == []
+
+
+def test_build_weekly_payload_flags_missing_data(tmp_path):
+    store = FakeStore()
+    store.quote_rows.append(quote_row(WEEK[0], "^GSPC", 102.0, 100.0))  # only Monday stored
+    payload, _ = build_weekly_payload(MONDAY, make_settings(tmp_path), FakeClock(), store)
+
+    assert any(g.startswith("weekly quotes: missing sessions") for g in payload.data_gaps)
+    assert "daily headlines: no stored briefings for last week" in payload.data_gaps
+    (q,) = payload.quotes
+    assert q.close == 102.0 and q.prev_close == 100.0  # single-session week still aggregates
+
+
+def test_build_weekly_payload_degraded_daily_report(tmp_path):
+    store = populated_store()
+    store.report_rows[0] = report_row(WEEK[0] + timedelta(days=1), WEEK[0], None)  # degraded day
+    payload, _ = build_weekly_payload(MONDAY, make_settings(tmp_path), FakeClock(), store)
+    assert payload.history_notes[0].endswith("(요약 없음)")
+
+
+# ------------------------------------------------------------------- orchestration
+class Env:
+    def __init__(self, monkeypatch, tmp_path, **settings_overrides):
+        self.settings = make_settings(tmp_path, **settings_overrides)
+        self.clock = FakeClock()
+        self.store = populated_store()
+        self.calls: dict[str, int] = {}
+        self.delivered: list = []
+
+        def count(name):
+            self.calls[name] = self.calls.get(name, 0) + 1
+
+        def summarize_weekly_payload(payload, sessions, provider):
+            count("llm")
+            assert sessions == WEEK
+            return sample_report(), "MiniMax-M3"
+
+        def verify_and_revise(payload, report, provider, *, data_message=None, system_prompt=None):
+            count("verify")
+            assert system_prompt == WEEKLY_SYSTEM_PROMPT
+            assert data_message and "## 주간 시세" in data_message
+            return report, Verification(ok=True)
+
+        def deliver(result, settings, store, report_id):
+            count("deliver")
+            self.delivered.append(result)
+            return []
+
+        monkeypatch.setattr(weekly, "MarketClock", lambda offset, **_kw: self.clock)
+        monkeypatch.setattr(weekly, "Store", lambda path: self.store)
+        monkeypatch.setattr(weekly, "build_provider", lambda settings: object())
+        monkeypatch.setattr(weekly, "summarize_weekly_payload", summarize_weekly_payload)
+        monkeypatch.setattr(weekly, "verify_and_revise", verify_and_revise)
+        monkeypatch.setattr(weekly, "_deliver", deliver)
+        monkeypatch.setattr(weekly, "_now", lambda: NOW)
+        self.monkeypatch = monkeypatch
+
+
+@pytest.fixture
+def env(monkeypatch, tmp_path):
+    return Env(monkeypatch, tmp_path)
+
+
+def test_run_weekly_happy_path(env):
+    result = run_weekly(MONDAY, env.settings)
+
+    assert result.kind == "weekly" and result.degraded is False and result.error is None
+    assert result.payload.krx_date == MONDAY
+    assert result.verification is not None and result.verification.ok is True
+    assert env.calls == {"llm": 1, "verify": 1, "deliver": 1}
+    assert env.store.reports == {(MONDAY, "weekly"): 1}
+    assert "미국 주간 리뷰" in result.markdown and "반도체 랠리로 주간 상승" in result.markdown
+    assert env.store.closed is True
+
+
+def test_run_weekly_defaults_to_latest_monday(env):
+    result = run_weekly(None, env.settings)
+    assert result.payload.krx_date == MONDAY  # NOW is that Monday morning
+
+
+def test_run_weekly_skips_when_already_sent(env):
+    env.store.reports[(MONDAY, "weekly")] = 99
+    result = run_weekly(MONDAY, env.settings)
+    assert result.error == SKIP_ALREADY_SENT
+    assert "deliver" not in env.calls
+
+
+def test_run_weekly_force_resends(env):
+    env.store.reports[(MONDAY, "weekly")] = 99
+    result = run_weekly(MONDAY, env.settings, force=True)
+    assert result.error is None and env.calls["deliver"] == 1
+
+
+def test_run_weekly_rejects_non_monday(env):
+    with pytest.raises(ValueError, match="not a Monday"):
+        run_weekly(date(2026, 9, 15), env.settings)
+
+
+def test_run_weekly_rejects_future_monday(env):
+    with pytest.raises(ValueError, match="future"):
+        run_weekly(date(2026, 9, 21), env.settings)
+
+
+def test_run_weekly_dry_run_prints_and_saves_nothing(env, capsys):
+    result = run_weekly(MONDAY, env.settings, dry_run=True)
+    assert "미국 주간 리뷰" in capsys.readouterr().out
+    assert env.store.reports == {} and "deliver" not in env.calls
+    assert result.kind == "weekly"
+
+
+def test_run_weekly_no_llm_degrades(env):
+    result = run_weekly(MONDAY, env.settings, no_llm=True)
+    assert result.degraded is True and result.report is None
+    assert "llm" not in env.calls and "verify" not in env.calls
+    assert env.calls["deliver"] == 1
+    assert "축소 모드" in result.markdown
+
+
+def test_run_weekly_llm_failure_degrades(env):
+    def boom(payload, sessions, provider):
+        raise LLMError("all models down")
+
+    env.monkeypatch.setattr(weekly, "summarize_weekly_payload", boom)
+    result = run_weekly(MONDAY, env.settings)
+    assert result.degraded is True and result.error == "all models down"
+    assert env.calls["deliver"] == 1
+
+
+def test_run_weekly_verify_failure_is_non_fatal(env):
+    def boom(payload, report, provider, **kwargs):
+        raise RuntimeError("verifier down")
+
+    env.monkeypatch.setattr(weekly, "verify_and_revise", boom)
+    result = run_weekly(MONDAY, env.settings)
+    assert result.report is not None and result.verification is None and result.error is None
+
+
+def test_run_weekly_verification_disabled(monkeypatch, tmp_path):
+    env = Env(monkeypatch, tmp_path, US2KR_VERIFY_ENABLED="false")
+    result = run_weekly(MONDAY, env.settings)
+    assert "verify" not in env.calls and result.verification is None
+
+
+# ------------------------------------------------------------------- prompt & render
+def test_build_weekly_user_message_contents(env, tmp_path):
+    payload, sessions = build_weekly_payload(MONDAY, env.settings, FakeClock(), populated_store())
+    message = build_weekly_user_message(payload, sessions)
+    assert "지난주 미국 거래일: 2026-09-07 ~ 2026-09-11 (5거래일)" in message
+    assert "## 주간 시세 (변동 = 주간 누적)" in message
+    assert "^GSPC S&P 500: 110.00 (+10.00%)" in message
+    assert "## 일별 요약" in message and "미 2026-09-08(화): 8일 헤드라인" in message
+
+
+def test_weekly_system_prompt_is_static_and_weekly_framed():
+    import re
+
+    assert "주간" in WEEKLY_SYSTEM_PROMPT and "미국 거래일" in WEEKLY_SYSTEM_PROMPT
+    assert not re.search(r"\d{4}-\d{2}-\d{2}", WEEKLY_SYSTEM_PROMPT)
+    assert not re.search(r"20\d{2}", WEEKLY_SYSTEM_PROMPT)
+
+
+def test_render_weekly_markdown(env):
+    payload, sessions = build_weekly_payload(MONDAY, env.settings, FakeClock(), populated_store())
+    md = render_weekly_markdown(payload, sample_report(), sessions)
+    assert md.startswith("# 🇺🇸→🇰🇷 미국 주간 리뷰 — 2026-09-14 (미 2026-09-07 ~ 2026-09-11)")
+    assert "## 주간 총평\n**반도체 랠리로 주간 상승**" in md
+    assert "## 이번 주 한국 장 시사점" in md and "## 이번 주 볼 것" in md
+    assert "📊 주요 지수 — 주간 변동" in md and "| S&P 500 | 110.00 | +10.00% |" in md
+    assert "## 일별 헤드라인" in md
+
+
+def test_render_weekly_markdown_degraded(env):
+    payload, sessions = build_weekly_payload(MONDAY, env.settings, FakeClock(), populated_store())
+    md = render_weekly_markdown(payload, None, sessions)
+    assert "🧩 축소 모드(LLM 요약 없음)" in md and "## 주간 총평" not in md
+
+
+# ------------------------------------------------------------------- daemon wiring
+from us2kr.daemon import WEEKLY_JOB_ID, Daemon
+
+
+class MiniScheduler:
+    def __init__(self):
+        self.jobs: dict = {}
+
+    def add_job(self, func, trigger=None, **kwargs):
+        self.jobs[kwargs.get("id")] = {"func": func, "trigger": trigger, **kwargs}
+
+    def add_listener(self, callback, mask=0):
+        pass
+
+    def shutdown(self, wait=True):
+        pass
+
+
+def make_daemon(tmp_path, *, now=NOW, weekly_fn=..., store=None, **settings_overrides):
+    calls: list[date] = []
+    if weekly_fn is ...:
+        weekly_fn = calls.append
+    daemon = Daemon(
+        make_settings(tmp_path, **settings_overrides),
+        scheduler=MiniScheduler(),
+        clock=FakeClock(),
+        store=store or FakeStore(),
+        run_fn=lambda session: None,
+        now_fn=lambda: now,
+        deliverers_fn=lambda s: [],
+        weekly_fn=weekly_fn,
+    )
+    return daemon, calls
+
+
+def test_daemon_schedules_weekly_cron(tmp_path):
+    daemon, _ = make_daemon(tmp_path)
+    daemon.schedule_weekly()
+    job = daemon.scheduler.jobs[WEEKLY_JOB_ID]
+    fields = {f.name: str(f) for f in job["trigger"].fields}
+    assert fields["day_of_week"] == "mon" and fields["hour"] == "7" and fields["minute"] == "0"
+
+
+def test_daemon_weekly_schedule_respects_disable_and_missing_fn(tmp_path):
+    daemon, _ = make_daemon(tmp_path, US2KR_WEEKLY_ENABLED="false")
+    daemon.schedule_weekly()
+    assert WEEKLY_JOB_ID not in daemon.scheduler.jobs
+
+    daemon, _ = make_daemon(tmp_path, weekly_fn=None)
+    daemon.schedule_weekly()
+    assert WEEKLY_JOB_ID not in daemon.scheduler.jobs
+
+
+def test_daemon_weekly_catchup_runs_missed_recap(tmp_path):
+    daemon, calls = make_daemon(tmp_path, now=datetime(2026, 9, 14, 8, 40, tzinfo=KST))
+    assert daemon.run_weekly_catchup() is True
+    assert calls == [MONDAY]
+
+
+def test_daemon_weekly_catchup_skips_when_sent_or_out_of_window(tmp_path):
+    store = FakeStore()
+    store.reports[(MONDAY, "weekly")] = 1
+    daemon, calls = make_daemon(tmp_path, now=datetime(2026, 9, 14, 8, 40, tzinfo=KST), store=store)
+    assert daemon.run_weekly_catchup() is False and calls == []
+
+    daemon, calls = make_daemon(tmp_path, now=datetime(2026, 9, 15, 8, 40, tzinfo=KST))  # Tuesday
+    assert daemon.run_weekly_catchup() is False and calls == []
+
+    daemon, calls = make_daemon(tmp_path, now=datetime(2026, 9, 14, 6, 30, tzinfo=KST))  # before 07:00
+    assert daemon.run_weekly_catchup() is False and calls == []
+
+    daemon, calls = make_daemon(tmp_path, now=datetime(2026, 9, 14, 10, 30, tzinfo=KST))  # window passed
+    assert daemon.run_weekly_catchup() is False and calls == []
+
+
+def test_daemon_weekly_job_failure_notifies_but_survives(tmp_path):
+    daemon, _ = make_daemon(tmp_path, now=datetime(2026, 9, 14, 7, 0, tzinfo=KST))
+    errors: list[tuple[str, str]] = []
+    daemon._notify_error = lambda subject, detail: errors.append((subject, detail))
+    daemon.weekly_fn = lambda monday: (_ for _ in ()).throw(RuntimeError("boom"))
+    daemon._weekly_job()  # must not raise
+    assert errors and "주간 리뷰" in errors[0][0]
diff --git a/us2kr/cli.py b/us2kr/cli.py
index fc0840f..9738b85 100644
--- a/us2kr/cli.py
+++ b/us2kr/cli.py
@@ -17,6 +17,7 @@ from us2kr.deliver import build_deliverers
 from us2kr.logging_setup import setup_logging
 from us2kr.market_time import MarketClock
 from us2kr.pipeline import SKIP_ALREADY_SENT, SKIP_KRX_CLOSED, delivery_failed, run_briefing
+from us2kr.weekly import run_weekly
 
 log = logging.getLogger(__name__)
 
@@ -77,6 +78,11 @@ def build_parser() -> argparse.ArgumentParser:
     once.add_argument("--dry-run", action="store_true", help="print markdown, do not save or deliver")
     once.add_argument("--force", action="store_true", help="re-run (resend) even if already sent")
     once.add_argument("--no-llm", action="store_true", help="skip the LLM (tables only)")
+    once.add_argument(
+        "--weekly",
+        action="store_true",
+        help="build the Monday weekly recap instead (--date must be a Monday KST; default: latest Monday)",
+    )
 
     sub.add_parser("daemon", help="run the resident scheduler")
     sub.add_parser("next-run", help="show the next scheduled briefing")
@@ -107,7 +113,8 @@ def build_parser() -> argparse.ArgumentParser:
 
 # ---------------------------------------------------------------------------- commands
 def cmd_run_once(args: argparse.Namespace, settings: Settings) -> int:
-    result = run_briefing(args.date, settings, dry_run=args.dry_run, force=args.force, no_llm=args.no_llm)
+    run = run_weekly if args.weekly else run_briefing
+    result = run(args.date, settings, dry_run=args.dry_run, force=args.force, no_llm=args.no_llm)
     if result.error in (SKIP_ALREADY_SENT, SKIP_KRX_CLOSED):
         print(f"skipped: {result.error} (KRX date {result.payload.krx_date})", file=sys.stderr)
         return EXIT_SKIPPED
diff --git a/us2kr/config.py b/us2kr/config.py
index 8561b79..4ccdffa 100644
--- a/us2kr/config.py
+++ b/us2kr/config.py
@@ -69,6 +69,9 @@ class Settings(BaseSettings):
     krx_extra_holidays: Annotated[list[date], NoDecode] = Field(default_factory=list, alias="US2KR_KRX_EXTRA_HOLIDAYS")
     premarket_update_enabled: bool = Field(default=False, alias="US2KR_PREMARKET_UPDATE_ENABLED")
     premarket_time_kst: str = Field(default="08:20", alias="US2KR_PREMARKET_TIME_KST")
+    # Monday-morning weekly recap (no US session closes into Monday KST, so the slot is free).
+    weekly_enabled: bool = Field(default=True, alias="US2KR_WEEKLY_ENABLED")
+    weekly_time_kst: str = Field(default="07:00", alias="US2KR_WEEKLY_TIME_KST")
 
     # --- Collection policy ---
     news_lookback_hours: int = Field(default=18, alias="US2KR_NEWS_LOOKBACK_HOURS")
diff --git a/us2kr/daemon.py b/us2kr/daemon.py
index c4be155..23be64b 100644
--- a/us2kr/daemon.py
+++ b/us2kr/daemon.py
@@ -22,12 +22,14 @@ from us2kr.deliver import build_deliverers
 from us2kr.market_time import MarketClock
 from us2kr.pipeline import run_briefing
 from us2kr.store import Store
+from us2kr.weekly import run_weekly
 
 log = logging.getLogger(__name__)
 
 BRIEFING_JOB_ID = "briefing"
 RECOVERY_JOB_ID = "briefing-recover"
 PREMARKET_JOB_ID = "premarket"
+WEEKLY_JOB_ID = "weekly"
 HEARTBEAT_JOB_ID = "heartbeat"
 
 # None => APScheduler runs the one-shot briefing job whenever it becomes due, even hours late (machine
@@ -64,6 +66,7 @@ class Daemon:
         now_fn: Callable[[], datetime] | None = None,
         deliverers_fn: Callable[[Settings], list[Any]] | None = None,
         premarket_fn: Callable[[], Any] | None = None,
+        weekly_fn: Callable[[date], Any] | None = None,
     ) -> None:
         self.settings = settings
         self.scheduler = scheduler
@@ -74,6 +77,8 @@ class Daemon:
         self.deliverers_fn = deliverers_fn or build_deliverers
         # Extension point: plug a real pre-market update here (PLAN.md P6).
         self.premarket_fn = premarket_fn
+        # Monday weekly recap; None disables (run_daemon wires us2kr.weekly.run_weekly).
+        self.weekly_fn = weekly_fn
         self.next_run: tuple[date, datetime] | None = None
         self.stop_requested = False
         # Serialises pipeline runs (scheduled job / catch-up / heartbeat recovery) across executor threads.
@@ -87,8 +92,13 @@ class Daemon:
         if self.stop_requested:  # signal arrived during catch-up, before the scheduler was running
             log.info("shutdown requested during catch-up - not scheduling jobs")
             return
+        self.run_weekly_catchup()
+        if self.stop_requested:
+            log.info("shutdown requested during weekly catch-up - not scheduling jobs")
+            return
         self.schedule_next()
         self.schedule_premarket()
+        self.schedule_weekly()
         self.schedule_heartbeat()
 
     def install_signal_handlers(self) -> None:
@@ -181,6 +191,46 @@ class Daemon:
         )
         log.info("premarket update scheduled Mon-Fri %02d:%02d KST", hour, minute)
 
+    def schedule_weekly(self) -> None:
+        if not (self.settings.weekly_enabled and self.weekly_fn):
+            return
+        hour, minute = _parse_hhmm(self.settings.weekly_time_kst)
+        self.scheduler.add_job(
+            self._weekly_job,
+            CronTrigger(day_of_week="mon", hour=hour, minute=minute, timezone=ZoneInfo(KST)),
+            id=WEEKLY_JOB_ID,
+            name="weekly recap",
+            replace_existing=True,
+            misfire_grace_time=3600,
+        )
+        log.info("weekly recap scheduled Mon %02d:%02d KST", hour, minute)
+
+    def run_weekly_catchup(self) -> bool:
+        """On a Monday restart, run a missed weekly recap within the catch-up window. True if run."""
+        if not (self.settings.weekly_enabled and self.weekly_fn):
+            return False
+        now = self.now_fn()
+        if now.weekday() != 0:
+            return False
+        hour, minute = _parse_hhmm(self.settings.weekly_time_kst)
+        due = now.replace(hour=hour, minute=minute, second=0, microsecond=0)
+        if not (due <= now < due + timedelta(hours=self.settings.catchup_window_hours)):
+            return False
+        if self.store.has_report(now.date(), kind="weekly"):
+            log.info("weekly catch-up: recap for %s already sent", now.date())
+            return False
+        if not self._run_lock.acquire(blocking=False):
+            return False
+        try:
+            log.info("weekly catch-up: running missed recap for %s", now.date())
+            self.weekly_fn(now.date())
+        except Exception:  # daemon must keep going
+            log.exception("weekly catch-up for %s failed", now.date())
+            self._notify_error("US2KR 주간 리뷰 실패", f"monday={now.date()}")
+        finally:
+            self._run_lock.release()
+        return True
+
     def schedule_heartbeat(self) -> None:
         self.scheduler.add_job(
             self._heartbeat_job,
@@ -224,6 +274,18 @@ class Daemon:
         except Exception:
             log.exception("premarket update failed")
 
+    def _weekly_job(self) -> None:
+        """Monday recap. run_weekly itself is idempotent per Monday (reports kind='weekly')."""
+        monday = self.now_fn().date()
+        try:
+            assert self.weekly_fn is not None  # scheduled only when set
+            log.info("scheduled weekly recap starting for %s", monday)
+            with self._run_lock:
+                self.weekly_fn(monday)
+        except Exception as exc:  # never let a job kill the scheduler
+            log.exception("weekly recap for %s failed", monday)
+            self._notify_error("US2KR 주간 리뷰 실패", f"monday={monday}: {type(exc).__name__}: {exc}"[:500])
+
     def _heartbeat_job(self) -> None:
         """Hourly liveness log; also the backstop that re-arms a missing or stale briefing schedule."""
         now = self.now_fn()
@@ -314,6 +376,7 @@ def run_daemon(settings: Settings) -> None:
         clock=clock,
         store=store,
         run_fn=lambda session: run_briefing(session, settings),
+        weekly_fn=lambda monday: run_weekly(monday, settings),
     )
     daemon.install_signal_handlers()
     log.info(
diff --git a/us2kr/deliver/blogger.py b/us2kr/deliver/blogger.py
index dfb8934..15d8e35 100644
--- a/us2kr/deliver/blogger.py
+++ b/us2kr/deliver/blogger.py
@@ -54,11 +54,12 @@ def briefing_html(markdown_text: str) -> str:
 
 def post_title_prefix(result: BriefingResult) -> str:
     """Stable per-day prefix - resends update the existing post instead of creating a duplicate."""
-    return f"미국 마감 브리핑 {result.payload.krx_date.isoformat()}"
+    label = "미국 주간 리뷰" if result.kind == "weekly" else "미국 마감 브리핑"
+    return f"{label} {result.payload.krx_date.isoformat()}"
 
 
 def post_title(result: BriefingResult) -> str:
-    """'미국 마감 브리핑 YYYY-MM-DD — 헤드라인' (falls back to the session date without a report)."""
+    """'{prefix} — 헤드라인' (falls back to the session date without a report)."""
     base = post_title_prefix(result)
     headline = (result.report.headline or "").strip() if result.report else ""
     return f"{base} — {headline}" if headline else f"{base} (미 {result.payload.session_date.isoformat()} 세션)"
diff --git a/us2kr/llm/base.py b/us2kr/llm/base.py
index 48652f5..0f8948b 100644
--- a/us2kr/llm/base.py
+++ b/us2kr/llm/base.py
@@ -16,12 +16,24 @@ class LLMProvider(Protocol):
         """Return (validated Report, model id that produced it). Raise LLMError on total failure."""
         ...
 
-    def verify(self, payload: MarketPayload, report: Report) -> tuple[Verification, str]:
-        """Fact-check `report` against the payload. Raise LLMError on total failure."""
+    def summarize_weekly(self, payload: MarketPayload, week_sessions: list) -> tuple[Report, str]:
+        """Weekly recap over a payload whose quotes carry week-over-week changes."""
+        ...
+
+    def verify(
+        self, payload: MarketPayload, report: Report, data_message: str | None = None
+    ) -> tuple[Verification, str]:
+        """Fact-check `report` against the payload (or `data_message`). Raise LLMError on total failure."""
         ...
 
     def revise(
-        self, payload: MarketPayload, report: Report, issues: list[VerificationIssue]
+        self,
+        payload: MarketPayload,
+        report: Report,
+        issues: list[VerificationIssue],
+        *,
+        system_prompt: str | None = None,
+        data_message: str | None = None,
     ) -> tuple[Report, str]:
         """Return a corrected Report addressing `issues`. Raise LLMError on total failure."""
         ...
diff --git a/us2kr/llm/minimax.py b/us2kr/llm/minimax.py
index 8898c2e..ca2fa9d 100644
--- a/us2kr/llm/minimax.py
+++ b/us2kr/llm/minimax.py
@@ -17,8 +17,10 @@ from us2kr.llm.prompts import (
     REVISE_INSTRUCTION,
     SYSTEM_PROMPT,
     VERIFY_SYSTEM_PROMPT,
+    WEEKLY_SYSTEM_PROMPT,
     build_user_message,
-    build_verify_message,
+    build_verify_message_from,
+    build_weekly_user_message,
     format_issues,
 )
 from us2kr.models import MarketPayload, Report, Verification, VerificationIssue
@@ -216,21 +218,38 @@ class MiniMaxProvider:
         ]
         return self._run(messages, Report, "report", task="summarize")
 
-    def verify(self, payload: MarketPayload, report: Report) -> tuple[Verification, str]:
-        """Fact-check `report` against the payload. Raise LLMError on total failure."""
+    def summarize_weekly(self, payload: MarketPayload, week_sessions: list) -> tuple[Report, str]:
+        """Weekly recap: payload.quotes carry week-over-week changes, history_notes the daily headlines."""
+        messages: list[dict[str, Any]] = [
+            {"role": "system", "content": WEEKLY_SYSTEM_PROMPT},
+            {"role": "user", "content": build_weekly_user_message(payload, week_sessions)},
+        ]
+        return self._run(messages, Report, "report", task="summarize-weekly")
+
+    def verify(
+        self, payload: MarketPayload, report: Report, data_message: str | None = None
+    ) -> tuple[Verification, str]:
+        """Fact-check `report` against the payload (or an explicit serialisation of it)."""
+        source = data_message if data_message is not None else build_user_message(payload)
         messages: list[dict[str, Any]] = [
             {"role": "system", "content": VERIFY_SYSTEM_PROMPT},
-            {"role": "user", "content": build_verify_message(payload, report)},
+            {"role": "user", "content": build_verify_message_from(source, report)},
         ]
         return self._run(messages, Verification, "verification", task="verify")
 
     def revise(
-        self, payload: MarketPayload, report: Report, issues: list[VerificationIssue]
+        self,
+        payload: MarketPayload,
+        report: Report,
+        issues: list[VerificationIssue],
+        *,
+        system_prompt: str | None = None,
+        data_message: str | None = None,
     ) -> tuple[Report, str]:
         """Ask for a corrected Report given verification issues. Raise LLMError on total failure."""
         messages: list[dict[str, Any]] = [
-            {"role": "system", "content": SYSTEM_PROMPT},
-            {"role": "user", "content": build_user_message(payload)},
+            {"role": "system", "content": system_prompt or SYSTEM_PROMPT},
+            {"role": "user", "content": data_message or build_user_message(payload)},
             {"role": "assistant", "content": report.model_dump_json()},
             {"role": "user", "content": REVISE_INSTRUCTION + format_issues(issues)},
         ]
diff --git a/us2kr/llm/prompts.py b/us2kr/llm/prompts.py
index 1074ecf..b23ef82 100644
--- a/us2kr/llm/prompts.py
+++ b/us2kr/llm/prompts.py
@@ -169,6 +169,33 @@ def _bullets(items: list[str]) -> list[str]:
     return [f"- {item}" for item in items] or ["(없음)"]
 
 
+WEEKLY_SYSTEM_PROMPT: str = f"""당신은 한국 투자자를 위한 미국 시장 주간 리뷰 작성자입니다.
+주말이 지난 월요일 아침, 한국 장이 열리기 전에 읽히는 "지난주 미국 시장 정리"를 작성합니다.
+
+## 작성 규칙
+1. 숫자는 사용자 메시지에 제공된 값만 사용합니다. 시세의 변동률은 **한 주 누적 변동**이므로 반드시
+   "주간" 관점으로 서술하고, 하루 변동처럼 쓰지 않습니다. 제공되지 않은 숫자를 만들거나 추정하지 않습니다.
+2. "일별 요약"은 그 주의 일일 브리핑 헤드라인입니다. 사건 서술은 이 범위 안에서만 하고, 헤드라인만으로
+   확인되지 않는 배경·인과는 "보도에 따르면"으로 표기하거나 생략합니다.
+3. 투자 조언(매수/매도/비중 조절 권고)을 하지 않습니다. 사실과 흐름을 정리하는 정보 요약 톤을 유지합니다.
+4. 한국 장 관점을 반드시 포함합니다: 반도체(SOX 등)·환율·한국 ADR/EWY의 주간 흐름과, 이번 주 한국 시장에
+   주는 시사점을 다룹니다.
+5. 간결한 한국어 경어체(~습니다)로 작성합니다. 모든 필드는 한국어로만 작성합니다(고유명사 예외).
+6. 날짜·요일·시점 표현은 전부 **미국 거래일** 기준입니다. 세션 정보에 주어진 주간 범위(첫 세션~마지막 세션)
+   밖의 시점을 서술하지 않습니다.
+7. 출력 스키마는 일일 브리핑과 같지만 의미를 주간으로 바꿔 씁니다:
+   headline=주간 총평 한 줄, market_overview=주간 시장 개요, key_drivers=지난주 주요 사건/요인,
+   rates_fx_commodities=금리·환율·원자재의 주간 흐름, korea_implications=이번 주 한국 장 시사점,
+   watchlist_today=이번 주 한국 장에서 확인할 것, risk_flags=이번 주 리스크.
+8. 출력은 지정된 JSON 스키마만 반환합니다. 배열 필드는 반드시 JSON 배열(각 항목 한 문장)로 출력합니다.
+
+## 입력 데이터 용어집
+시세는 아래 그룹으로 나뉘어 제공되며, 각 줄의 형식은 `심볼 이름: 주말 종가 (주간 변동)` 입니다.
+국채 금리의 변동은 주간 수익률 변화폭(%p)입니다. 값이 없는 항목은 "자료 없음"으로 표시됩니다.
+{_GROUP_GLOSSARY}
+"""
+
+
 VERIFY_SYSTEM_PROMPT: str = """당신은 미국 시장 브리핑의 팩트체커입니다. 브리핑 초안이 함께 제공된 원본 데이터와
 모순되지 않는지만 검사합니다. 문체·구성·의견은 평가하지 않습니다.
 
@@ -195,14 +222,14 @@ REVISE_INSTRUCTION: str = (
 )
 
 
+def build_verify_message_from(data_message: str, report: Report) -> str:
+    """Fact-check user message from an explicit data serialisation plus the draft report."""
+    return data_message + "\n## 검증 대상 브리핑 초안 (JSON)\n" + report.model_dump_json(indent=2) + "\n"
+
+
 def build_verify_message(payload: MarketPayload, report: Report) -> str:
     """User message for the fact-check pass: the payload serialisation plus the draft report."""
-    return (
-        build_user_message(payload)
-        + "\n## 검증 대상 브리핑 초안 (JSON)\n"
-        + report.model_dump_json(indent=2)
-        + "\n"
-    )
+    return build_verify_message_from(build_user_message(payload), report)
 
 
 def format_issues(issues: list) -> str:
@@ -213,6 +240,36 @@ def format_issues(issues: list) -> str:
     return "\n".join(lines) or "(없음)"
 
 
+def build_weekly_user_message(payload: MarketPayload, week_sessions: list) -> str:
+    """Serialise a weekly payload (quotes = week-over-week changes, notes = daily headlines)."""
+    out: list[str] = []
+    first = week_sessions[0].isoformat() if week_sessions else "?"
+    last = week_sessions[-1].isoformat() if week_sessions else "?"
+
+    out.append("## 세션 정보 (주간)")
+    out.append(f"- 지난주 미국 거래일: {first} ~ {last} ({len(week_sessions)}거래일)")
+    out.append(f"- 한국 거래일(발행일): {payload.krx_date.isoformat()} (월요일)")
+    out.append(f"- 오늘 한국 장: {'개장' if payload.krx_open_today else '휴장'}")
+    out.append("")
+
+    out.append("## 주간 시세 (변동 = 주간 누적)")
+    by_group = payload.quotes_by_group()
+    if not by_group:
+        out.append("(시세 자료 없음)")
+    for group in _group_order(list(by_group)):
+        out.append(f"### {GROUP_TITLES_KO.get(group, group)}")
+        out.extend(_fmt_quote(q) for q in by_group[group])
+    out.append("")
+
+    out.append("## 일별 요약 (그 주의 일일 브리핑 헤드라인)")
+    out.extend(_bullets(payload.history_notes))
+    out.append("")
+
+    out.append("## 수집 누락")
+    out.extend(_bullets(payload.data_gaps))
+    return "\n".join(out) + "\n"
+
+
 def build_user_message(payload: MarketPayload) -> str:
     """Serialise the payload into a deterministic Korean text block for the user message."""
     out: list[str] = []
diff --git a/us2kr/models.py b/us2kr/models.py
index d936e85..c630b1f 100644
--- a/us2kr/models.py
+++ b/us2kr/models.py
@@ -118,3 +118,6 @@ class BriefingResult(BaseModel):
     degraded: bool = False  # True when LLM failed and only the numeric table was produced
     error: str | None = None
     verification: Verification | None = None  # None: not run (disabled / degraded / verifier failed)
+    # "briefing": daily post-close briefing. "weekly": Monday-morning recap, where payload.quotes carry
+    # week-over-week changes, payload.session_date is the week's last US session and krx_date the Monday.
+    kind: Literal["briefing", "weekly"] = "briefing"
diff --git a/us2kr/render.py b/us2kr/render.py
index 5829c18..9b32f70 100644
--- a/us2kr/render.py
+++ b/us2kr/render.py
@@ -221,6 +221,45 @@ def render_markdown(payload: MarketPayload, report: Report | None, model_used: s
     return "\n".join(lines).rstrip() + "\n"
 
 
+# ------------------------------------------------------------------ weekly recap
+
+
+def _weekly_report_sections(report: Report) -> list[str]:
+    out: list[str] = []
+    if _text(report.headline):
+        out += _section("주간 총평", [f"**{_text(report.headline)}**"])
+    if _text(report.market_overview):
+        out += _section("주간 개요", [_text(report.market_overview)])
+    out += _section("지난주 주요 사건", _bullets(report.key_drivers))
+    if _text(report.rates_fx_commodities):
+        out += _section("금리·환율·원자재", [_text(report.rates_fx_commodities)])
+    out += _section("이번 주 한국 장 시사점", _bullets(report.korea_implications))
+    out += _section("이번 주 볼 것", _bullets(report.watchlist_today))
+    out += _section("리스크", _bullets(report.risk_flags))
+    return out
+
+
+def render_weekly_markdown(payload: MarketPayload, report: Report | None, week_sessions: list) -> str:
+    """Render the Monday weekly recap. `payload.quotes` carry week-over-week changes."""
+    first = week_sessions[0].isoformat() if week_sessions else "?"
+    last = week_sessions[-1].isoformat() if week_sessions else "?"
+    lines = [f"# 🇺🇸→🇰🇷 미국 주간 리뷰 — {payload.krx_date.isoformat()} (미 {first} ~ {last})"]
+    gen_kst = _kst(payload.generated_at).strftime("%Y-%m-%d %H:%M")
+    lines.append(f"지난주 미국 거래일 {len(week_sessions)}일 · 생성: {gen_kst} KST")
+    if not payload.krx_open_today:
+        lines.append("⚠️ 오늘 한국 휴장")
+    if report is None:
+        lines.append("🧩 축소 모드(LLM 요약 없음)")
+    lines.append("")
+    if report is not None:
+        lines += _weekly_report_sections(report)
+    for g, title, qs in _ordered_groups(payload):
+        lines += _section(f"📊 {title} — 주간 변동", _quote_table(g, qs))
+    lines += _section("일별 헤드라인", _bullets(payload.history_notes))
+    lines += _section("수집 누락", _bullets(payload.data_gaps))
+    return "\n".join(lines).rstrip() + "\n"
+
+
 # ------------------------------------------------------------------ Telegram HTML
 
 _LINK_RE = re.compile(r"\[([^\]]+)\]\(([^)\s]+)\)")
diff --git a/us2kr/store.py b/us2kr/store.py
index cb1c339..c3929db 100644
--- a/us2kr/store.py
+++ b/us2kr/store.py
@@ -146,6 +146,17 @@ class Store:
         )
         return list(cur.fetchall())
 
+    def load_quotes_between(self, start: date, end: date) -> list[sqlite3.Row]:
+        """Quote rows for every stored session in [start, end], ordered by (symbol, session_date)."""
+        cur = self._conn.execute(
+            """
+            SELECT * FROM quotes WHERE session_date >= ? AND session_date <= ?
+            ORDER BY symbol, session_date
+            """,
+            (start.isoformat(), end.isoformat()),
+        )
+        return list(cur.fetchall())
+
     def _history(self, symbol: str, before: date, n: int) -> list[_Row]:
         """Up to ``n`` stored rows for ``symbol`` strictly before ``before``, oldest first."""
         cur = self._conn.execute(
@@ -250,6 +261,17 @@ class Store:
         log.info("saved report id=%d krx_date=%s kind=%s degraded=%s", report_id, krx, kind, result.degraded)
         return report_id
 
+    def reports_between(self, start: date, end: date, kind: str = "briefing") -> list[sqlite3.Row]:
+        """(krx_date, session_date, result_json) rows with start <= krx_date <= end, oldest first."""
+        cur = self._conn.execute(
+            """
+            SELECT krx_date, session_date, result_json FROM reports
+            WHERE kind = ? AND krx_date >= ? AND krx_date <= ? ORDER BY krx_date
+            """,
+            (kind, start.isoformat(), end.isoformat()),
+        )
+        return list(cur.fetchall())
+
     def mark_sent(self, report_id: int, channel: str, receipt: str) -> None:
         with self._conn:
             self._conn.execute(
diff --git a/us2kr/summarize.py b/us2kr/summarize.py
index ff83bde..30a18a3 100644
--- a/us2kr/summarize.py
+++ b/us2kr/summarize.py
@@ -17,16 +17,32 @@ def build_provider(settings: Settings) -> LLMProvider:
     return MiniMaxProvider(settings)
 
 
-def summarize_payload(payload: MarketPayload, provider: LLMProvider) -> tuple[Report, str]:
-    """Summarise via the provider and guarantee `report.data_gaps` includes every payload gap."""
-    report, model_used = provider.summarize(payload)
+def _merge_gaps(payload: MarketPayload, report: Report) -> Report:
+    """Guarantee `report.data_gaps` includes every payload gap (deduplicated, payload first)."""
     merged: list[str] = []
     for gap in [*payload.data_gaps, *report.data_gaps]:
         if gap and gap not in merged:
             merged.append(gap)
     if merged != list(report.data_gaps):
         report = report.model_copy(update={"data_gaps": merged})
-    log.info("summary produced by %s (%d data gaps)", model_used, len(merged))
+    return report
+
+
+def summarize_payload(payload: MarketPayload, provider: LLMProvider) -> tuple[Report, str]:
+    """Summarise via the provider and guarantee `report.data_gaps` includes every payload gap."""
+    report, model_used = provider.summarize(payload)
+    report = _merge_gaps(payload, report)
+    log.info("summary produced by %s (%d data gaps)", model_used, len(report.data_gaps))
+    return report, model_used
+
+
+def summarize_weekly_payload(
+    payload: MarketPayload, week_sessions: list, provider: LLMProvider
+) -> tuple[Report, str]:
+    """Weekly-recap variant of summarize_payload."""
+    report, model_used = provider.summarize_weekly(payload, week_sessions)
+    report = _merge_gaps(payload, report)
+    log.info("weekly summary produced by %s (%d data gaps)", model_used, len(report.data_gaps))
     return report, model_used
 
 
@@ -37,14 +53,21 @@ def _consistent(verification: Verification) -> Verification:
 
 
 def verify_and_revise(
-    payload: MarketPayload, report: Report, provider: LLMProvider
+    payload: MarketPayload,
+    report: Report,
+    provider: LLMProvider,
+    *,
+    data_message: str | None = None,
+    system_prompt: str | None = None,
 ) -> tuple[Report, Verification]:
     """Fact-check the report; on major issues ask for one revision and re-check it.
 
-    Returns (possibly revised report, verification of the returned report). Raises LLMError only when
-    the *initial* verification cannot be produced; a failed revision keeps the original report.
+    `data_message`/`system_prompt` override the default daily-briefing serialisation and author persona
+    (used by the weekly recap). Returns (possibly revised report, verification of the returned report).
+    Raises LLMError only when the *initial* verification cannot be produced; a failed revision keeps
+    the original report.
     """
-    raw_verification, model = provider.verify(payload, report)
+    raw_verification, model = provider.verify(payload, report, data_message)
     verification = _consistent(raw_verification)
     log.info(
         "verification by %s: ok=%s issues=%d (major=%s)",
@@ -59,9 +82,11 @@ def verify_and_revise(
     for issue in verification.issues:
         log.warning("verification issue [%s] %s -> %s", issue.severity, issue.claim, issue.problem)
     try:
-        revised, model = provider.revise(payload, report, verification.issues)
+        revised, model = provider.revise(
+            payload, report, verification.issues, system_prompt=system_prompt, data_message=data_message
+        )
         log.info("report revised by %s after verification issues", model)
-        re_verification, model = provider.verify(payload, revised)
+        re_verification, model = provider.verify(payload, revised, data_message)
         re_verification = _consistent(re_verification)
         log.info(
             "re-verification by %s: ok=%s issues=%d", model, re_verification.ok, len(re_verification.issues)
diff --git a/us2kr/weekly.py b/us2kr/weekly.py
new file mode 100644
index 0000000..81b7402
--- /dev/null
+++ b/us2kr/weekly.py
@@ -0,0 +1,242 @@
+"""Monday-morning weekly recap: aggregate the stored week into one briefing.
+
+No US session closes into Monday KST (the US Sunday never trades), so the Monday slot carries a
+recap of the previous week's US sessions instead. Everything is built from what the daily pipeline
+already stored - the `quotes` table for week-over-week changes and the daily reports for the
+narrative - so no collector runs.
+
+The result is a regular ``BriefingResult`` with ``kind="weekly"`` whose payload reuses
+``MarketPayload``: ``quotes`` carry weekly changes (close = Friday close, prev_close = the previous
+week's close), ``history_notes`` carry the daily briefing headlines, ``session_date`` is the week's
+last US session and ``krx_date`` the Monday. That keeps the store, every deliverer and the
+dashboard working unchanged.
+"""
+
+from __future__ import annotations
+
+import json
+import logging
+from datetime import date, datetime, timedelta
+from zoneinfo import ZoneInfo
+
+from us2kr.config import KST, Settings
+from us2kr.llm.base import LLMError
+from us2kr.llm.prompts import WEEKLY_SYSTEM_PROMPT, build_weekly_user_message
+from us2kr.market_time import MarketClock
+from us2kr.models import BriefingResult, MarketPayload, Quote, Report, Verification
+from us2kr.pipeline import SKIP_ALREADY_SENT, _deliver, _skip_result
+from us2kr.render import render_weekly_markdown
+from us2kr.store import Store
+from us2kr.summarize import build_provider, summarize_weekly_payload, verify_and_revise
+from us2kr.symbols import display_name, group_of
+
+log = logging.getLogger(__name__)
+
+WEEKDAY_KO = ("월", "화", "수", "목", "금", "토", "일")
+
+
+def _now() -> datetime:
+    return datetime.now(ZoneInfo(KST))
+
+
+def most_recent_monday(today: date) -> date:
+    """Today if it is a Monday, otherwise the latest past Monday."""
+    return today - timedelta(days=today.weekday())
+
+
+def week_sessions_for(monday: date, clock: MarketClock) -> list[date]:
+    """The previous week's US sessions (Mon..Fri before `monday`), holidays excluded."""
+    return [
+        d
+        for d in (monday - timedelta(days=offset) for offset in range(7, 2, -1))
+        if clock.is_us_session(d)
+    ]
+
+
+def _weekly_quote(symbol: str, rows: list) -> Quote:
+    """Collapse one symbol's session rows (oldest first) into a week-over-week Quote."""
+    first, last = rows[0], rows[-1]
+    base = first["prev_close"] if first["prev_close"] is not None else first["close"]
+    end = last["close"]
+    change_pct = None
+    change_abs = None
+    if end is not None and base:
+        change_pct = (end / base - 1.0) * 100.0
+        change_abs = end - base
+    highs = [r["day_high"] for r in rows if r["day_high"] is not None]
+    lows = [r["day_low"] for r in rows if r["day_low"] is not None]
+    return Quote(
+        symbol=symbol,
+        name=display_name(symbol),
+        group=group_of(symbol) or "etc",
+        close=end,
+        prev_close=base,
+        change_pct=change_pct,
+        change_abs=change_abs,
+        day_high=max(highs) if highs else None,
+        day_low=min(lows) if lows else None,
+        source="store",
+    )
+
+
+def _daily_headline_notes(store: Store, monday: date, gaps: list[str]) -> list[str]:
+    """'미 {세션}({요일}): {헤드라인}' lines from the week's stored daily briefings."""
+    notes: list[str] = []
+    for row in store.reports_between(monday - timedelta(days=6), monday - timedelta(days=1)):
+        try:
+            data = json.loads(row["result_json"] or "{}")
+        except json.JSONDecodeError:
+            gaps.append(f"daily report {row['krx_date']}: unreadable")
+            continue
+        session = row["session_date"] or data.get("payload", {}).get("session_date") or "?"
+        try:
+            weekday = WEEKDAY_KO[date.fromisoformat(session).weekday()]
+            label = f"미 {session}({weekday})"
+        except ValueError:
+            label = f"미 {session}"
+        report = data.get("report") or {}
+        headline = (report.get("headline") or "").strip()
+        notes.append(f"{label}: {headline}" if headline else f"{label}: (요약 없음)")
+    return notes
+
+
+def build_weekly_payload(
+    monday: date, settings: Settings, clock: MarketClock, store: Store
+) -> tuple[MarketPayload, list[date]]:
+    """Assemble the weekly MarketPayload for the Monday `monday` from stored data only."""
+    sessions = week_sessions_for(monday, clock)
+    gaps: list[str] = []
+
+    quotes: list[Quote] = []
+    covered: set[str] = set()
+    if sessions:
+        by_symbol: dict[str, list] = {}
+        for row in store.load_quotes_between(sessions[0], sessions[-1]):
+            by_symbol.setdefault(row["symbol"], []).append(row)
+            covered.add(row["session_date"])
+        quotes = [_weekly_quote(symbol, rows) for symbol, rows in sorted(by_symbol.items())]
+    missing = [s for s in sessions if s.isoformat() not in covered]
+    if not quotes:
+        gaps.append("weekly quotes: no stored sessions for last week")
+    elif missing:
+        gaps.append("weekly quotes: missing sessions " + ", ".join(s.isoformat() for s in missing))
+
+    history_notes = _daily_headline_notes(store, monday, gaps)
+    if not history_notes:
+        gaps.append("daily headlines: no stored briefings for last week")
+
+    last_session = sessions[-1] if sessions else monday - timedelta(days=3)
+    payload = MarketPayload(
+        session_date=last_session,
+        us_close_time=clock.us_session_close(last_session),
+        generated_at=_now(),
+        krx_date=monday,
+        krx_open_today=clock.is_krx_session(monday),
+        quotes=quotes,
+        history_notes=history_notes,
+        data_gaps=gaps,
+    )
+    log.info(
+        "weekly payload for %s: %d sessions, %d symbols, %d headlines, %d gaps",
+        monday,
+        len(sessions),
+        len(quotes),
+        len(history_notes),
+        len(gaps),
+    )
+    return payload, sessions
+
+
+def _summarize_weekly(
+    payload: MarketPayload, sessions: list[date], settings: Settings
+) -> tuple[Report | None, str | None, str | None]:
+    """Return (report, model_used, error); any failure degrades instead of aborting (pipeline policy)."""
+    try:
+        provider = build_provider(settings)
+        report, model_used = summarize_weekly_payload(payload, sessions, provider)
+        return report, model_used, None
+    except LLMError as exc:
+        log.error("weekly LLM summarisation failed, degrading: %s", exc)
+        return None, None, str(exc)
+    except Exception as exc:
+        log.exception("unexpected weekly LLM failure, degrading")
+        return None, None, f"{type(exc).__name__}: {exc}"
+
+
+def _verify_weekly(
+    payload: MarketPayload, sessions: list[date], report: Report, settings: Settings
+) -> tuple[Report, Verification | None]:
+    try:
+        provider = build_provider(settings)
+        return verify_and_revise(
+            payload,
+            report,
+            provider,
+            data_message=build_weekly_user_message(payload, sessions),
+            system_prompt=WEEKLY_SYSTEM_PROMPT,
+        )
+    except Exception as exc:  # noqa: BLE001 - verification must never block the recap
+        log.warning("weekly verification failed, delivering unverified: %s: %s", type(exc).__name__, exc)
+        return report, None
+
+
+def run_weekly(
+    monday: date | None,
+    settings: Settings,
+    *,
+    dry_run: bool = False,
+    force: bool = False,
+    no_llm: bool = False,
+) -> BriefingResult:
+    """Produce and deliver the weekly recap for ``monday`` (default: the most recent Monday, KST)."""
+    clock = MarketClock(settings.run_offset_minutes, extra_krx_holidays=settings.krx_extra_holidays)
+    store = Store(settings.db_path)
+    try:
+        now = _now()
+        if monday is None:
+            monday = most_recent_monday(now.date())
+        if monday.weekday() != 0:
+            raise ValueError(f"{monday} is not a Monday - the weekly recap is keyed by Mondays (KST)")
+        if monday > now.date():
+            raise ValueError(f"{monday} is in the future")
+
+        payload, sessions = build_weekly_payload(monday, settings, clock, store)
+        if not dry_run and not force and store.has_report(monday, kind="weekly"):
+            log.info("weekly recap for %s already sent - skipping (use --force to resend)", monday)
+            return _skip_result(payload, SKIP_ALREADY_SENT)
+
+        report: Report | None = None
+        model_used: str | None = None
+        error: str | None = None
+        verification: Verification | None = None
+        degraded = False
+        if no_llm:
+            degraded = True
+            log.info("--no-llm: skipping weekly summarisation")
+        else:
+            report, model_used, error = _summarize_weekly(payload, sessions, settings)
+            degraded = report is None
+            if report is not None and settings.verify_enabled:
+                report, verification = _verify_weekly(payload, sessions, report, settings)
+
+        markdown = render_weekly_markdown(payload, report, sessions)
+        result = BriefingResult(
+            payload=payload,
+            report=report,
+            markdown=markdown,
+            model_used=model_used,
+            degraded=degraded,
+            error=error,
+            verification=verification,
+            kind="weekly",
+        )
+        if dry_run:
+            print(markdown)
+            return result
+
+        report_id = store.save_report(result, kind="weekly")
+        log.info("weekly report %d saved for %s (degraded=%s)", report_id, monday, degraded)
+        _deliver(result, settings, store, report_id)
+        return result
+    finally:
+        store.close()