diff --git a/app/core/config.py b/app/core/config.py
index 8d7ec14..e699582 100644
--- a/app/core/config.py
+++ b/app/core/config.py
@@ -157,6 +157,8 @@ class Settings(BaseModel):
     # PR-MacMini-Derived-Worker-1: study explanation owner = Mac mini
     # GPU 측은 false 로 설정 (.env), explanation 분기 skip guard 트리거.
     study_explanation_enabled: bool = True
+    # 공부 암기노트 Phase 1: card_extract 폴러/consumer 게이트. owner 분리 시 false 로.
+    study_card_extract_enabled: bool = True
 
     # internal endpoint Bearer token (Mac mini derived-worker 호출용)
     internal_worker_token: str = ""
@@ -167,6 +169,7 @@ def load_settings() -> Settings:
     # 환경변수 (docker-compose에서 주입)
     database_url = os.getenv("DATABASE_URL", "")
     study_explanation_enabled = os.getenv("STUDY_EXPLANATION_ENABLED", "true").lower() in ("1", "true", "yes")
+    study_card_extract_enabled = os.getenv("STUDY_CARD_EXTRACT_ENABLED", "true").lower() in ("1", "true", "yes")
     internal_worker_token = os.getenv("INTERNAL_WORKER_TOKEN", "")
     jwt_secret = os.getenv("JWT_SECRET", "")
     totp_secret = os.getenv("TOTP_SECRET", "")
@@ -262,6 +265,7 @@ def load_settings() -> Settings:
         document_types=document_types,
         upload=upload_cfg,
         study_explanation_enabled=study_explanation_enabled,
+        study_card_extract_enabled=study_card_extract_enabled,
         internal_worker_token=internal_worker_token,
     )
 
diff --git a/app/main.py b/app/main.py
index 68c380b..c4861f2 100644
--- a/app/main.py
+++ b/app/main.py
@@ -54,6 +54,8 @@ async def lifespan(app: FastAPI):
     from workers.queue_consumer import consume_queue, consume_markdown_queue
     from workers.study_queue_consumer import consume_study_queue
     from workers.study_session_queue_consumer import consume_study_session_queue
+    from workers.study_memo_card_jobs_consumer import consume_study_memo_card_queue
+    from workers.study_card_enqueue import run as study_card_enqueue_run
     from workers.study_question_embed_worker import (
         refresh_stale_related as study_q_related_refresh,
         run as study_q_embed_run,
@@ -95,6 +97,10 @@ async def lifespan(app: FastAPI):
     # Phase 4-B v1: study_quiz_session_jobs 처리 — 세션 단위 자유 마크다운 분석.
     # 4-A 와 같은 MLX gate 공유 — 4-A 처리 중이면 직렬 대기.
     scheduler.add_job(consume_study_session_queue, "interval", minutes=1, id="study_session_queue_consumer")
+    # 공부 암기노트 Phase 1: card_extract 큐 consumer + 버전키 폴러(study_card_enqueue).
+    # 별 테이블/별 consumer 로 기존 study queue 와 격리. settings.study_card_extract_enabled 게이트.
+    scheduler.add_job(consume_study_memo_card_queue, "interval", minutes=1, id="study_memo_card_consumer")
+    scheduler.add_job(study_card_enqueue_run, "interval", minutes=1, id="study_card_enqueue")
     # PR-B 레거시 tier 백필 — 30분 주기로 호출되지만 KST 00:00~06:00 시간대만 실제 enqueue.
     # safety > law > manual 우선순위로 25건씩. 6720 레거시 → 야간당 ~150건 → 약 45일 소화.
     scheduler.add_job(tier_backfill_run, "interval", minutes=30, id="tier_backfill")
diff --git a/app/models/study_memo_card.py b/app/models/study_memo_card.py
new file mode 100644
index 0000000..0551a63
--- /dev/null
+++ b/app/models/study_memo_card.py
@@ -0,0 +1,211 @@
+"""study_memo_cards / study_memo_card_evidence ORM (공부 암기노트 Phase 1).
+
+study_questions(MCQ) 와 별개로, 풀이/근거에서 추출한 암기 플래시카드 본체.
+  - source_kind: question(P1) / subject_note / document(P3 예약)
+  - format: qa(cue->fact) / cloze(빈칸). 강한 enum 미사용 (read-time 매핑).
+  - source_generated_at: 추출 당시 ai_explanation_generated_at — 버전 핀/stale 판정.
+  - needs_review DEFAULT true: 생성물이라 검토 대기로 입고.
+
+dedup_hash PARTIAL UNIQUE(migration 288, WHERE deleted_at IS NULL) 가 중복 최종 방어선.
+정정/삭제 후 supersede(구버전 카드 deleted_at 마킹)로 stale 잔류 0 — append 전에 호출해
+살아있는 구카드가 새 추출을 ON CONFLICT 로 막지 않게 한다.
+"""
+
+from __future__ import annotations
+
+from datetime import datetime
+from typing import Any, Sequence
+
+from sqlalchemy import (
+    BigInteger,
+    Boolean,
+    DateTime,
+    ForeignKey,
+    Integer,
+    String,
+    Text,
+    func,
+    text,
+    update,
+)
+from sqlalchemy.dialects.postgresql import JSONB, insert as pg_insert
+from sqlalchemy.ext.asyncio import AsyncSession
+from sqlalchemy.orm import Mapped, mapped_column
+
+from core.database import Base
+
+
+class StudyMemoCard(Base):
+    __tablename__ = "study_memo_cards"
+
+    id: Mapped[int] = mapped_column(BigInteger, primary_key=True)
+    user_id: Mapped[int] = mapped_column(
+        BigInteger, ForeignKey("users.id", ondelete="CASCADE"), nullable=False
+    )
+    study_topic_id: Mapped[int] = mapped_column(
+        BigInteger, ForeignKey("study_topics.id", ondelete="CASCADE"), nullable=False
+    )
+
+    source_kind: Mapped[str] = mapped_column(String(40), nullable=False)
+    source_question_id: Mapped[int | None] = mapped_column(
+        BigInteger, ForeignKey("study_questions.id", ondelete="CASCADE")
+    )
+    source_subject_note_id: Mapped[int | None] = mapped_column(BigInteger)
+
+    format: Mapped[str] = mapped_column(String(20), nullable=False)
+    cue: Mapped[str] = mapped_column(Text, nullable=False)
+    fact: Mapped[str] = mapped_column(Text, nullable=False)
+    cloze_text: Mapped[str | None] = mapped_column(Text)
+    extra: Mapped[dict | None] = mapped_column(JSONB)
+
+    source_generated_at: Mapped[datetime | None] = mapped_column(DateTime(timezone=True))
+    dedup_hash: Mapped[str] = mapped_column(String(64), nullable=False)
+
+    needs_review: Mapped[bool] = mapped_column(Boolean, nullable=False, default=True)
+    flagged_at: Mapped[datetime | None] = mapped_column(DateTime(timezone=True))
+    flagged_by: Mapped[str | None] = mapped_column(String(40))
+
+    model: Mapped[str | None] = mapped_column(String(120))
+    generated_at: Mapped[datetime | None] = mapped_column(DateTime(timezone=True))
+    created_at: Mapped[datetime] = mapped_column(
+        DateTime(timezone=True), default=datetime.now, nullable=False
+    )
+    deleted_at: Mapped[datetime | None] = mapped_column(DateTime(timezone=True))
+
+
+class StudyMemoCardEvidence(Base):
+    """append-only citation. UPDATE/DELETE 없음."""
+
+    __tablename__ = "study_memo_card_evidence"
+
+    id: Mapped[int] = mapped_column(BigInteger, primary_key=True)
+    card_id: Mapped[int] = mapped_column(
+        BigInteger, ForeignKey("study_memo_cards.id", ondelete="CASCADE"), nullable=False
+    )
+    source_type: Mapped[str] = mapped_column(String(40), nullable=False)
+    source_id: Mapped[int | None] = mapped_column(BigInteger)
+    chunk_index: Mapped[int | None] = mapped_column(Integer)
+    snippet: Mapped[str | None] = mapped_column(Text)
+    created_at: Mapped[datetime] = mapped_column(
+        DateTime(timezone=True), default=datetime.now, nullable=False
+    )
+
+
+async def supersede_old_cards(
+    session: AsyncSession,
+    *,
+    source_question_id: int,
+    keep_generated_at: datetime | None,
+) -> int:
+    """같은 문제의 '다른 버전' 카드를 deleted_at 마킹(retire).
+
+    새 source_generated_at 카드 적재 '전에' 호출 — 살아있는 구버전 카드가 dedup PARTIAL
+    UNIQUE 로 새 추출을 막는 것을 방지(정정-후 stale 잔류 0). 같은 버전은 보존.
+    Returns: retire 된 행 수.
+    """
+    stmt = (
+        update(StudyMemoCard)
+        .where(
+            StudyMemoCard.source_question_id == source_question_id,
+            StudyMemoCard.deleted_at.is_(None),
+            StudyMemoCard.source_generated_at.is_distinct_from(keep_generated_at),
+        )
+        .values(deleted_at=func.now())
+    )
+    result = await session.execute(stmt)
+    return result.rowcount or 0
+
+
+async def append_card(
+    session: AsyncSession,
+    *,
+    user_id: int,
+    study_topic_id: int,
+    source_kind: str,
+    source_question_id: int | None,
+    format: str,
+    cue: str,
+    fact: str,
+    cloze_text: str | None,
+    dedup_hash: str,
+    source_generated_at: datetime | None,
+    model: str | None,
+    generated_at: datetime | None,
+    needs_review: bool = True,
+) -> int | None:
+    """카드 1장 INSERT. dedup_hash PARTIAL UNIQUE 충돌 시 None (DO NOTHING).
+
+    Returns: 새 card.id, 또는 중복으로 건너뛰면 None.
+    """
+    stmt = (
+        pg_insert(StudyMemoCard)
+        .values(
+            user_id=user_id,
+            study_topic_id=study_topic_id,
+            source_kind=source_kind,
+            source_question_id=source_question_id,
+            format=format,
+            cue=cue,
+            fact=fact,
+            cloze_text=cloze_text,
+            dedup_hash=dedup_hash,
+            source_generated_at=source_generated_at,
+            needs_review=needs_review,
+            model=model,
+            generated_at=generated_at,
+        )
+        .on_conflict_do_nothing(
+            index_elements=["dedup_hash"],
+            index_where=text("deleted_at IS NULL"),
+        )
+        .returning(StudyMemoCard.id)
+    )
+    result = await session.execute(stmt)
+    return result.scalar_one_or_none()
+
+
+async def append_card_evidence(
+    session: AsyncSession,
+    *,
+    card_id: int,
+    refs: Sequence[dict[str, Any]],
+) -> int:
+    """카드 인용 append-only INSERT. refs: [{source_type, source_id?, chunk_index?, snippet?}]."""
+    rows = [
+        {
+            "card_id": card_id,
+            "source_type": r.get("source_type") or "unknown",
+            "source_id": r.get("source_id"),
+            "chunk_index": r.get("chunk_index"),
+            "snippet": r.get("snippet"),
+        }
+        for r in refs
+    ]
+    if not rows:
+        return 0
+    await session.execute(pg_insert(StudyMemoCardEvidence).values(rows))
+    return len(rows)
+
+
+async def flag_cards_for_source(
+    session: AsyncSession,
+    *,
+    source_question_id: int,
+    reason: str,
+) -> int:
+    """소스 문제 정정/삭제 시 파생 카드를 needs_review=auto 마킹(임시 플래그).
+
+    최종 stale 정리는 워커 supersede 가 책임 — 이건 사용자 가시화용 즉시 플래그.
+    reason: 'source_changed' | 'source_deleted'.
+    Returns: 마킹된 행 수.
+    """
+    stmt = (
+        update(StudyMemoCard)
+        .where(
+            StudyMemoCard.source_question_id == source_question_id,
+            StudyMemoCard.deleted_at.is_(None),
+        )
+        .values(needs_review=True, flagged_by=reason, flagged_at=func.now())
+    )
+    result = await session.execute(stmt)
+    return result.rowcount or 0
diff --git a/app/models/study_memo_card_job.py b/app/models/study_memo_card_job.py
new file mode 100644
index 0000000..8bfe818
--- /dev/null
+++ b/app/models/study_memo_card_job.py
@@ -0,0 +1,92 @@
+"""study_memo_card_jobs ORM — card_extract 비동기 작업 큐 (다형 소스).
+
+231_study_question_jobs 복제 + source_kind/source_id/source_version(=ai_explanation_generated_at).
+별도 테이블 + 별도 consumer(study_memo_card_jobs_consumer.py) 로 기존 study_queue_consumer 와 격리.
+
+error_code 권장값:
+  - parse_fail / llm_timeout / unknown  → 재시도 대상 (attempts < max_attempts)
+  - all_dropped  → 0장 생성. completed 로 종결해 같은 버전 재추출 차단.
+  - no_ready_explanation  → ai_explanation 미준비(race). skipped, 비재시도.
+
+멱등 이중구조: active partial unique(migration 292)는 동시 active 1행만,
+버전 멱등(같은 source_version 재추출 차단)은 폴러의 NOT EXISTS(source_version) 가 책임.
+"""
+
+from __future__ import annotations
+
+from datetime import datetime
+from typing import Any
+
+from sqlalchemy import BigInteger, DateTime, ForeignKey, SmallInteger, String, Text, text
+from sqlalchemy.dialects.postgresql import JSONB, insert as pg_insert
+from sqlalchemy.ext.asyncio import AsyncSession
+from sqlalchemy.orm import Mapped, mapped_column
+
+from core.database import Base
+
+
+class StudyMemoCardJob(Base):
+    __tablename__ = "study_memo_card_jobs"
+
+    id: Mapped[int] = mapped_column(BigInteger, primary_key=True)
+    user_id: Mapped[int] = mapped_column(
+        BigInteger, ForeignKey("users.id", ondelete="CASCADE"), nullable=False
+    )
+
+    source_kind: Mapped[str] = mapped_column(String(40), nullable=False)
+    source_id: Mapped[int] = mapped_column(BigInteger, nullable=False)
+    source_version: Mapped[datetime | None] = mapped_column(DateTime(timezone=True))
+
+    kind: Mapped[str] = mapped_column(String(40), nullable=False)
+    status: Mapped[str] = mapped_column(String(20), nullable=False, default="pending")
+    attempts: Mapped[int] = mapped_column(SmallInteger, nullable=False, default=0)
+    max_attempts: Mapped[int] = mapped_column(SmallInteger, nullable=False, default=2)
+    error_code: Mapped[str | None] = mapped_column(String(40))
+    error_message: Mapped[str | None] = mapped_column(Text)
+    payload: Mapped[dict | None] = mapped_column(JSONB)
+    created_at: Mapped[datetime] = mapped_column(
+        DateTime(timezone=True), default=datetime.now, nullable=False
+    )
+    started_at: Mapped[datetime | None] = mapped_column(DateTime(timezone=True))
+    completed_at: Mapped[datetime | None] = mapped_column(DateTime(timezone=True))
+
+    # active partial unique idx (source_kind, source_id) WHERE active 는 migration 292.
+
+
+async def enqueue_study_memo_card_job(
+    session: AsyncSession,
+    *,
+    user_id: int,
+    source_kind: str,
+    source_id: int,
+    source_version: datetime | None,
+    kind: str = "card_extract",
+    payload: dict[str, Any] | None = None,
+) -> bool:
+    """study_memo_card_jobs 에 행 추가 (DB 레벨 동시 active 중복 방어).
+
+    같은 (source_kind, source_id) 활성 행(pending/processing)이 있으면 False.
+    버전 멱등(같은 source_version 재추출 차단)은 호출 측 폴러의 NOT EXISTS 가 선판단.
+
+    Returns: True = 새 enqueue, False = active 중복으로 건너뜀.
+    """
+    values: dict[str, Any] = {
+        "user_id": user_id,
+        "source_kind": source_kind,
+        "source_id": source_id,
+        "source_version": source_version,
+        "kind": kind,
+        "status": "pending",
+    }
+    if payload is not None:
+        values["payload"] = payload
+    stmt = (
+        pg_insert(StudyMemoCardJob)
+        .values(**values)
+        .on_conflict_do_nothing(
+            index_elements=["source_kind", "source_id"],
+            index_where=text("status IN ('pending', 'processing')"),
+        )
+    )
+    result = await session.execute(stmt)
+    return result.rowcount > 0
diff --git a/app/models/study_question.py b/app/models/study_question.py
index afb7f56..040fab1 100644
--- a/app/models/study_question.py
+++ b/app/models/study_question.py
@@ -80,6 +80,12 @@ class StudyQuestion(Base):
     related_computed_at: Mapped[datetime | None] = mapped_column(DateTime(timezone=True))
     related_threshold_version: Mapped[str | None] = mapped_column(String(20))
 
+    # 공부 암기노트 Phase 1: 검수 대기 플래그 (DDL=migration 296). 정정/삭제 훅 + needs_review 큐가 set/clear.
+    # flagged_by 권장값: 'user' / 'source_changed' / 'source_deleted' (서버측 상수, read-time 매핑).
+    needs_review: Mapped[bool] = mapped_column(Boolean, default=False, nullable=False)
+    flagged_at: Mapped[datetime | None] = mapped_column(DateTime(timezone=True))
+    flagged_by: Mapped[str | None] = mapped_column(String(40))
+
     created_at: Mapped[datetime] = mapped_column(
         DateTime(timezone=True), default=datetime.now, nullable=False
     )
diff --git a/app/prompts/study_card_envelope.txt b/app/prompts/study_card_envelope.txt
new file mode 100644
index 0000000..32756d4
--- /dev/null
+++ b/app/prompts/study_card_envelope.txt
@@ -0,0 +1,39 @@
+당신은 한국 기사시험(가스기사·산업안전기사 등) 필기 학습 보조 AI 입니다.
+이미 검증된 풀이와 근거 자료에서 '암기 플래시카드'를 추출합니다.
+
+【문제】
+{question_text}
+
+【보기】
+1. {choice_1}
+2. {choice_2}
+3. {choice_3}
+4. {choice_4}
+
+【사용자가 입력한 정답】
+{correct_choice}번
+
+【확정 풀이 (검증 통과, 정성 사실의 1순위 근거)】
+{ai_explanation}
+
+【참고 자료 (정량 cloze 의 원문 근거)】
+
+▼ 자료
+{documents_evidence_block}
+
+▼ 같은 주제의 다른 문제
+{questions_evidence_block}
+
+【카드 추출 지침】
+1. 위 '확정 풀이'와 '참고 자료'에서 시험에 나올 핵심 사실을 1~3장의 카드로 추출한다.
+2. 카드 형식(format)은 두 가지:
+   - "qa": cue(질문/단서) -> fact(핵심 사실 한 줄).
+   - "cloze": 완전한 사실 문장에서 핵심 토큰 하나를 빈칸 [____] 로 가린 cloze_text + 그 가린 정답을 fact 에.
+3. **정량 토큰(수치·압력·온도·기준값·표준번호·조항)을 cloze 정답으로 쓸 때, 그 토큰은 반드시 위 '참고 자료' 원문에 그대로 등장해야 한다.** 확정 풀이에만 있고 자료에 없는 수치는 카드로 만들지 않는다. 단위는 자료 표기 그대로 쓰고 환산하지 않는다.
+4. cue 에 정답(fact)을 노출하지 않는다. cloze_text 의 빈칸 밖 평문에도 정답을 노출하지 않는다.
+5. **할루시네이션 방지 (절대 규칙)**: 근거 없는 수치·공식·표준 번호·법령 조항을 새로 만들어내지 않는다. 자료/풀이에서 확인되지 않는 내용은 카드로 만들지 않는다. "보통 ~이다" 같은 모호한 단정도 근거 없으면 쓰지 않는다.
+6. 카드는 최대 3장. 가장 시험가치 높은 사실 위주로, 억지로 채우지 않는다(0장도 허용).
+7. **출력은 raw JSON 한 객체만**. 메타 설명·인사·코드 펜스·thinking 텍스트 없이.
+
+【출력 형식】
+{{"cards": [{{"format": "qa|cloze", "cue": "<앞면 단서/질문>", "fact": "<핵심 사실/정답 토큰>", "cloze_text": "<cloze 일 때만, 빈칸 [____] 포함 문장>"}}]}}
diff --git a/app/services/study/card_normalize.py b/app/services/study/card_normalize.py
new file mode 100644
index 0000000..5bfe5af
--- /dev/null
+++ b/app/services/study/card_normalize.py
@@ -0,0 +1,85 @@
+"""공부 암기노트 카드 — 정량 토큰 정규화 + dedup 키 + 누출/근거 1차 primitives.
+
+정규화 정책(보수적 = restrictive):
+  - NFC 유니코드 정규화
+  - 수치와 단위 사이 공백 제거 ('0.5 MPa' -> '0.5MPa')
+  - 천단위 구분자(콤마) 제거 ('1,000kg' -> '1000kg'), 숫자 3자리 그룹 한정
+  - 단위 환산 절대 금지 (원문 표기 보존 — LLM 오변환을 정규화로 흡수하지 않음)
+
+대소문자는 보존한다 (MPa vs mpa 는 다른 단위라 lowercase 안 함).
+dedup_hash = sha256(source_question_id | format | normalize_token(정답토큰)).
+"""
+
+from __future__ import annotations
+
+import hashlib
+import re
+import unicodedata
+
+# 수치 다음의 공백 + (단위로 시작하는) 토큰 사이 공백 제거.
+_NUM_UNIT_SPACE = re.compile(r"(\d)\s+(?=[A-Za-z℃°%‰Ωµμ/])")
+# 천단위 콤마: 숫자 뒤 콤마 + 정확히 3자리 숫자 그룹이 이어질 때만 (소수점/일반 콤마 보호).
+_THOUSANDS = re.compile(r"(?<=\d),(?=\d{3}(?:\D|$))")
+_WS = re.compile(r"\s+")
+# cloze 빈칸 마커: [____] / [___] / {{...}} / ____ 등.
+_BLANK = re.compile(r"\[_+\]|\{\{[^}]*\}\}|_{2,}")
+_DIGIT = re.compile(r"\d")
+
+
+def normalize_token(s: str | None) -> str:
+    """단일 정답 토큰 정규화 (대소문자 보존). dedup 키·근거 매칭의 단위."""
+    if not s:
+        return ""
+    s = unicodedata.normalize("NFC", s)
+    s = _NUM_UNIT_SPACE.sub(r"\1", s)
+    s = _THOUSANDS.sub("", s)
+    return s.strip()
+
+
+def normalize_for_match(s: str | None) -> str:
+    """근거 텍스트/문장 비교용 — 토큰 정규화 + 공백 축약 (대소문자 보존)."""
+    if not s:
+        return ""
+    s = normalize_token(s)
+    return _WS.sub(" ", s).strip()
+
+
+def compute_dedup_hash(source_question_id: int | None, fmt: str, answer_token: str | None) -> str:
+    """정본 키: sha256(source_question_id | format | normalize_token(정답토큰))."""
+    key = f"{source_question_id}|{fmt}|{normalize_token(answer_token)}"
+    return hashlib.sha256(key.encode("utf-8")).hexdigest()
+
+
+def is_quantitative(token: str | None) -> bool:
+    """숫자를 포함하면 정량 토큰 (정량 cloze 는 evidence 원문 등장 필수)."""
+    return bool(_DIGIT.search(normalize_token(token)))
+
+
+def text_contains(haystack: str | None, needle: str | None) -> bool:
+    """needle(정답토큰)이 haystack 안에 정규화 후 부분문자열로 등장하면 True."""
+    n = normalize_for_match(needle)
+    if not n:
+        return False
+    return n in normalize_for_match(haystack)
+
+
+def is_cue_leak(cue: str | None, answer_token: str | None) -> bool:
+    """cue(앞면)에 정답토큰이 노출되면 True (drop 대상)."""
+    return text_contains(cue, answer_token)
+
+
+def is_cloze_self_leak(cloze_text: str | None, answer_token: str | None) -> bool:
+    """cloze_text 의 빈칸 마커를 제거한 평문에 정답토큰이 노출되면 True (drop 대상)."""
+    if not cloze_text:
+        return False
+    stripped = _BLANK.sub(" ", cloze_text)
+    return text_contains(stripped, answer_token)
+
+
+def matching_evidence(answer_token: str | None, evidence_refs: list[dict]) -> list[dict]:
+    """정답토큰이 snippet 에 등장하는 evidence_refs 만 반환 (citation 적재용)."""
+    out: list[dict] = []
+    for ref in evidence_refs or []:
+        if text_contains(ref.get("snippet"), answer_token):
+            out.append(ref)
+    return out
diff --git a/app/services/study/study_memo_card_guards.py b/app/services/study/study_memo_card_guards.py
new file mode 100644
index 0000000..3360655
--- /dev/null
+++ b/app/services/study/study_memo_card_guards.py
@@ -0,0 +1,105 @@
+"""공부 암기노트 카드별 가드 — 추출된 카드 1장 검증 파이프라인.
+
+explanation 워커의 단일 answer_choice 환각가드를 카드 배열로 확장한다. 가드 4종:
+  1. 형식 유효성 — format in {qa, cloze}, cue/fact 비공백, cloze 는 cloze_text + 빈칸 마커 필요.
+  2. 근거(hallucination) — 정답토큰(fact)이 신뢰 텍스트에 등장해야 채택.
+       정량 토큰(숫자 포함): evidence 원문 snippet 에 등장 필수 (평문화된 ai_explanation 만으론 불충분).
+       비정량(개념): ai_explanation 또는 evidence snippet 에 등장.
+  3. 누출 — cue 에 정답 노출 / cloze 평문에 정답 노출 시 drop.
+  4. dedup — (source_question_id, format, normalize(정답토큰)) hash. 배치 내 중복 1장.
+
+무결성은 구조로(메모리 규칙): dedup_hash PARTIAL UNIQUE(migration 288)가 DB 최종 방어선,
+본 가드는 1차. 전부 drop 이면 빈 리스트 → 워커가 all_dropped 로 종결.
+"""
+
+from __future__ import annotations
+
+from dataclasses import dataclass, field
+
+from services.study import card_normalize as cn
+
+_VALID_FORMATS = {"qa", "cloze"}
+
+
+@dataclass
+class GuardedCard:
+    format: str
+    cue: str
+    fact: str
+    cloze_text: str | None
+    dedup_hash: str
+    matched_evidence: list[dict] = field(default_factory=list)
+
+
+def guard_card(
+    card: dict,
+    *,
+    source_question_id: int | None,
+    ai_explanation: str | None,
+    evidence_refs: list[dict],
+) -> GuardedCard | None:
+    """카드 1장 검증. 통과하면 GuardedCard, 탈락하면 None."""
+    fmt = (card.get("format") or "").strip()
+    cue = (card.get("cue") or "").strip()
+    fact = (card.get("fact") or "").strip()
+    cloze_text = card.get("cloze_text")
+    cloze_text = cloze_text.strip() if isinstance(cloze_text, str) else None
+
+    # 1. 형식 유효성
+    if fmt not in _VALID_FORMATS or not cue or not fact:
+        return None
+    if fmt == "cloze":
+        if not cloze_text or not cn._BLANK.search(cloze_text):
+            return None
+
+    # 3. 누출 (정답 노출)
+    if cn.is_cue_leak(cue, fact):
+        return None
+    if fmt == "cloze" and cn.is_cloze_self_leak(cloze_text, fact):
+        return None
+
+    # 2. 근거 (hallucination 차단)
+    matched = cn.matching_evidence(fact, evidence_refs)
+    if cn.is_quantitative(fact):
+        # 정량 토큰은 evidence 원문 등장 필수
+        if not matched:
+            return None
+    else:
+        # 비정량은 ai_explanation 또는 evidence 에 등장
+        if not matched and not cn.text_contains(ai_explanation, fact):
+            return None
+
+    return GuardedCard(
+        format=fmt,
+        cue=cue,
+        fact=fact,
+        cloze_text=cloze_text if fmt == "cloze" else None,
+        dedup_hash=cn.compute_dedup_hash(source_question_id, fmt, fact),
+        matched_evidence=matched,
+    )
+
+
+def guard_cards(
+    cards: list[dict],
+    *,
+    source_question_id: int | None,
+    ai_explanation: str | None,
+    evidence_refs: list[dict],
+) -> list[GuardedCard]:
+    """카드 배열 검증 + 배치 내 dedup_hash 중복 1장. 통과 카드만 반환."""
+    out: list[GuardedCard] = []
+    seen: set[str] = set()
+    for card in cards or []:
+        if not isinstance(card, dict):
+            continue
+        g = guard_card(
+            card,
+            source_question_id=source_question_id,
+            ai_explanation=ai_explanation,
+            evidence_refs=evidence_refs,
+        )
+        if g is None or g.dedup_hash in seen:
+            continue
+        seen.add(g.dedup_hash)
+        out.append(g)
+    return out
diff --git a/app/workers/study_card_enqueue.py b/app/workers/study_card_enqueue.py
new file mode 100644
index 0000000..972bc26
--- /dev/null
+++ b/app/workers/study_card_enqueue.py
@@ -0,0 +1,80 @@
+"""study_card_enqueue — 버전키 폴러 (공부 암기노트 Phase 1).
+
+ready ai_explanation 인데 '현재 버전' card_extract job 이 없는 question 을 enqueue.
+버전 멱등(핵심): NOT EXISTS(job WHERE source_kind='question' AND source_id=q.id
+                            AND source_version=q.ai_explanation_generated_at)
+  - 같은 버전 재추출 차단 — completed/all_dropped job 도 현 버전에 존재하면 재enqueue 0(livelock 방지).
+  - explanation 재생성(새 generated_at)이면 새 버전 job 부재 → 자동 재추출(정정-stale 해소).
+NULL 가드: ai_explanation_generated_at IS NOT NULL 전제 — NULL 이면 NULL=NULL=UNKNOWN 으로
+  NOT EXISTS 가 항상 참이 되어 매 폴 재enqueue 폭주. ready 전이 직후 race 를 이 가드가 막는다.
+thundering-herd: per-poll LIMIT(CARD_ENQUEUE_BATCH) + 최근(generated_at desc) 우선으로 backfill 완만.
+"""
+
+from __future__ import annotations
+
+import logging
+
+from sqlalchemy import select
+
+from core.config import settings
+from core.database import async_session
+from models.study_memo_card_job import StudyMemoCardJob, enqueue_study_memo_card_job
+from models.study_question import StudyQuestion
+
+logger = logging.getLogger("study_card_enqueue")
+
+CARD_ENQUEUE_BATCH = 20
+SOURCE_KIND_QUESTION = "question"
+
+
+async def run() -> None:
+    """APScheduler 진입점. ready & 현 버전 job 부재 question 을 BATCH 만큼 enqueue."""
+    if not getattr(settings, "study_card_extract_enabled", True):
+        return
+
+    async with async_session() as session:
+        # 현재 ai_explanation_generated_at 버전에 대한 job 이 이미 있는지 (correlated NOT EXISTS).
+        job_exists = (
+            select(StudyMemoCardJob.id)
+            .where(
+                StudyMemoCardJob.source_kind == SOURCE_KIND_QUESTION,
+                StudyMemoCardJob.source_id == StudyQuestion.id,
+                StudyMemoCardJob.source_version == StudyQuestion.ai_explanation_generated_at,
+            )
+            .exists()
+        )
+        rows = (
+            await session.execute(
+                select(
+                    StudyQuestion.id,
+                    StudyQuestion.user_id,
+                    StudyQuestion.ai_explanation_generated_at,
+                )
+                .where(
+                    StudyQuestion.deleted_at.is_(None),
+                    StudyQuestion.ai_explanation_status == "ready",
+                    StudyQuestion.ai_explanation_generated_at.is_not(None),
+                    ~job_exists,
+                )
+                .order_by(StudyQuestion.ai_explanation_generated_at.desc())
+                .limit(CARD_ENQUEUE_BATCH)
+            )
+        ).all()
+        if not rows:
+            return
+
+        enqueued = 0
+        for r in rows:
+            ok = await enqueue_study_memo_card_job(
+                session,
+                user_id=r.user_id,
+                source_kind=SOURCE_KIND_QUESTION,
+                source_id=r.id,
+                source_version=r.ai_explanation_generated_at,
+                kind="card_extract",
+            )
+            if ok:
+                enqueued += 1
+        await session.commit()
+        if enqueued:
+            logger.info("study_card_enqueue candidates=%d enqueued=%d", len(rows), enqueued)
diff --git a/app/workers/study_memo_card_jobs_consumer.py b/app/workers/study_memo_card_jobs_consumer.py
new file mode 100644
index 0000000..9386bd7
--- /dev/null
+++ b/app/workers/study_memo_card_jobs_consumer.py
@@ -0,0 +1,87 @@
+"""study_memo_card_jobs consumer — APScheduler 1분 간격 (공부 암기노트 Phase 1).
+
+기존 study_queue_consumer / study_session_queue_consumer 와 별도 테이블·별도 consumer 라
+자연 격리된다 (정본 제약: 기존 consumer 무수정). study_session_queue_consumer 골격 복제.
+  BATCH_SIZE=1, MLX gate Semaphore(1) 공유 — explanation/session 처리 중이면 직렬 대기.
+  STALE_MINUTES=10 자체 복구.
+dispatch: kind=='card_extract' -> run_card_extract_job, 그 외 -> skipped(else 분기 silent loss 방지).
+"""
+
+from __future__ import annotations
+
+from datetime import datetime, timedelta, timezone
+
+from sqlalchemy import select, update
+from sqlalchemy.exc import SQLAlchemyError
+
+from core.database import async_session
+from core.utils import setup_logger
+from models.study_memo_card_job import StudyMemoCardJob
+from workers.study_memo_card_worker import run_card_extract_job
+
+logger = setup_logger("study_memo_card_jobs_consumer")
+
+BATCH_SIZE = 1
+STALE_MINUTES = 10
+
+
+async def reset_stale_card_jobs() -> None:
+    """processing 으로 STALE_MINUTES 이상 방치된 job 을 pending 으로 복구."""
+    cutoff = datetime.now(timezone.utc) - timedelta(minutes=STALE_MINUTES)
+    try:
+        async with async_session() as session:
+            stmt = (
+                update(StudyMemoCardJob)
+                .where(
+                    StudyMemoCardJob.status == "processing",
+                    StudyMemoCardJob.started_at.is_not(None),
+                    StudyMemoCardJob.started_at < cutoff,
+                )
+                .values(status="pending", started_at=None)
+            )
+            result = await session.execute(stmt)
+            await session.commit()
+            n = result.rowcount or 0
+            if n > 0:
+                logger.warning("study_memo_card_jobs_stale_reset count=%s", n)
+    except SQLAlchemyError as e:
+        logger.exception("study_memo_card_jobs_stale_reset_failed: %s", e)
+
+
+async def consume_study_memo_card_queue() -> None:
+    """APScheduler 진입점. pending card_extract job 을 BATCH_SIZE 만큼 처리."""
+    await reset_stale_card_jobs()
+
+    async with async_session() as session:
+        rows = (
+            await session.execute(
+                select(StudyMemoCardJob)
+                .where(StudyMemoCardJob.status == "pending")
+                .order_by(StudyMemoCardJob.id.asc())
+                .limit(BATCH_SIZE)
+            )
+        ).scalars().all()
+
+    for job_row in rows:
+        async with async_session() as s:
+            try:
+                job = await s.get(StudyMemoCardJob, job_row.id)
+                if job is None or job.status != "pending":
+                    continue
+                if job.kind == "card_extract":
+                    await run_card_extract_job(s, job)
+                else:
+                    # 미지원 kind — lost-in-queue 방지 위해 명시 skipped.
+                    job.status = "skipped"
+                    job.error_code = "unknown"
+                    job.error_message = f"unsupported kind: {job.kind!r}"
+                    job.completed_at = datetime.now(timezone.utc)
+                await s.commit()
+                logger.info(
+                    "card_extract_processed id=%s src=%s/%s status=%s error_code=%s attempts=%s",
+                    job.id, job.source_kind, job.source_id, job.status, job.error_code,
+                    job.attempts,
+                )
+            except Exception as e:
+                await s.rollback()
+                logger.exception("card_extract_outer_failed job_id=%s: %s", job_row.id, e)
diff --git a/app/workers/study_memo_card_worker.py b/app/workers/study_memo_card_worker.py
new file mode 100644
index 0000000..791458c
--- /dev/null
+++ b/app/workers/study_memo_card_worker.py
@@ -0,0 +1,251 @@
+"""card_extract 워커 — ready 풀이/근거에서 암기 플래시카드 추출 (공부 암기노트 Phase 1).
+
+study_explanation_worker.run_explanation_job 골격 복제:
+  1. ready 게이트 + RAG(gather_explanation_context) + ai_explanation + evidence_refs (in-process)
+  2. ai_explanation 미준비면 LLM 호출 X (no_ready_explanation, skipped)
+  3. MLX primary (gate BACKGROUND, gate 안에서만 timeout) -> {cards:[...]} JSON
+  4. 카드별 가드(study_memo_card_guards) — 근거(정량=원문등장)·누출·dedup
+  5. 통과 카드 있으면 supersede(구버전 retire) -> append + evidence. 0장이면 all_dropped(completed).
+
+GPU = in-process RAG provider (explanation 워커와 동일 구조; internal_study card-context
+endpoint 는 호출자 0 scaffold). Mac mini = call_primary 생성.
+재시도: llm_timeout/parse_fail/unknown 만 (attempts < max_attempts), all_dropped/no_ready 는 종결.
+"""
+
+from __future__ import annotations
+
+import asyncio
+import json
+import logging
+from datetime import datetime, timezone
+from pathlib import Path
+
+import httpx
+from sqlalchemy.ext.asyncio import AsyncSession
+
+from ai.client import AIClient, parse_json_response
+from models.study_memo_card import (
+    append_card,
+    append_card_evidence,
+    supersede_old_cards,
+)
+from models.study_memo_card_job import StudyMemoCardJob
+from models.study_question import StudyQuestion
+from models.user import User  # noqa: F401  (mapper 초기화 defensive)
+from services.search.llm_gate import Priority, acquire_mlx_gate
+from services.study.explanation_rag import (
+    gather_explanation_context,
+    render_evidence_block,
+)
+from services.study.study_memo_card_guards import guard_cards
+
+logger = logging.getLogger("study_memo_card_worker")
+
+# 다카드 출력이라 explanation(30s)보다 여유. config primary.timeout(180, soft-lock)은 미변경.
+CARD_LLM_TIMEOUT_S = 45.0
+SOURCE_KIND_QUESTION = "question"
+
+_ENVELOPE_PROMPT_FILE = "study_card_envelope.txt"
+_envelope_template_cache: str | None = None
+
+
+def _load_card_envelope_prompt() -> str:
+    global _envelope_template_cache
+    if _envelope_template_cache is None:
+        prompts_dir = Path(__file__).resolve().parent.parent / "prompts"
+        _envelope_template_cache = (
+            prompts_dir / _ENVELOPE_PROMPT_FILE
+        ).read_text(encoding="utf-8")
+    return _envelope_template_cache
+
+
+def _render_card_envelope_prompt(
+    q: StudyQuestion, doc_block: str, q_block: str, ai_explanation: str
+) -> str:
+    return (
+        _load_card_envelope_prompt()
+        .replace("{question_text}", q.question_text or "")
+        .replace("{choice_1}", q.choice_1 or "")
+        .replace("{choice_2}", q.choice_2 or "")
+        .replace("{choice_3}", q.choice_3 or "")
+        .replace("{choice_4}", q.choice_4 or "")
+        .replace("{correct_choice}", str(q.correct_choice))
+        .replace("{ai_explanation}", ai_explanation or "")
+        .replace("{documents_evidence_block}", doc_block)
+        .replace("{questions_evidence_block}", q_block)
+    )
+
+
+async def run_card_extract_job(session: AsyncSession, job: StudyMemoCardJob) -> None:
+    """study_memo_card_jobs row 1건 처리. caller(consumer)가 commit 책임.
+
+    종료 시 completed / failed / skipped / pending(재시도) 중 하나.
+    """
+    now = lambda: datetime.now(timezone.utc)  # noqa: E731
+
+    job.attempts += 1
+    job.status = "processing"
+    job.started_at = now()
+    await session.flush()
+
+    try:
+        # P1 은 question source 만. 다른 source_kind 는 미구현 — skipped.
+        if job.source_kind != SOURCE_KIND_QUESTION:
+            job.error_code = "unknown"
+            job.error_message = f"unsupported source_kind: {job.source_kind!r}"
+            job.status = "skipped"
+            job.completed_at = now()
+            return
+
+        question = await session.get(StudyQuestion, job.source_id)
+        if question is None or question.deleted_at is not None:
+            job.error_code = "no_ready_explanation"
+            job.error_message = "source question deleted or missing"
+            job.status = "skipped"
+            job.completed_at = now()
+            return
+
+        # ready 게이트 — explanation 이 ready 가 아니면(정정으로 stale 등) 추출 보류.
+        if question.ai_explanation_status != "ready" or not (question.ai_explanation or "").strip():
+            job.error_code = "no_ready_explanation"
+            job.error_message = f"ai_explanation_status={question.ai_explanation_status}"
+            job.status = "skipped"
+            job.completed_at = now()
+            return
+
+        source_version = question.ai_explanation_generated_at
+
+        # 1. RAG 근거 (in-process). ai_explanation 이 정성 1순위, evidence 가 정량 원문.
+        ctx = await gather_explanation_context(session, question.user_id, question)
+        evidence_refs = [it.to_dict() for it in ctx.all]
+        doc_block = render_evidence_block(ctx.documents)
+        q_block = render_evidence_block(ctx.questions)
+        prompt = _render_card_envelope_prompt(question, doc_block, q_block, question.ai_explanation)
+
+        # 2. MLX primary
+        ai_client = AIClient()
+        try:
+            async with acquire_mlx_gate(Priority.BACKGROUND):
+                async with asyncio.timeout(CARD_LLM_TIMEOUT_S):
+                    raw_text = await ai_client.call_primary(prompt)
+            primary_name = (
+                ai_client.ai.primary.model
+                if hasattr(ai_client.ai, "primary") and hasattr(ai_client.ai.primary, "model")
+                else "primary"
+            )
+        finally:
+            await ai_client.close()
+
+        if not raw_text or not raw_text.strip():
+            job.error_code = "llm_timeout"
+            job.error_message = "empty response from primary"
+            return
+
+        # 3. {cards:[...]} 파싱
+        def _save_raw_preview(reason: str) -> None:
+            existing = dict(job.payload or {})
+            existing["debug_raw_preview"] = (raw_text or "")[:1000]
+            existing["parse_fail_reason"] = reason
+            job.payload = existing
+
+        envelope = parse_json_response(raw_text)
+        if envelope is None or not isinstance(envelope, dict):
+            job.error_code = "parse_fail"
+            job.error_message = "envelope JSON parse failed"
+            _save_raw_preview("not_dict")
+            return
+        cards = envelope.get("cards")
+        if not isinstance(cards, list):
+            job.error_code = "parse_fail"
+            job.error_message = f"cards not a list: {type(cards).__name__}"
+            _save_raw_preview("cards_not_list")
+            return
+
+        # 4. 카드별 가드
+        guarded = guard_cards(
+            cards,
+            source_question_id=question.id,
+            ai_explanation=question.ai_explanation,
+            evidence_refs=evidence_refs,
+        )
+
+        payload = dict(job.payload or {})
+        payload["cards_generated"] = len(cards)
+        payload["cards_kept"] = len(guarded)
+
+        if not guarded:
+            # 전량 drop — completed 로 종결해 같은 버전 재추출 차단(재시도 집합에서 제외).
+            payload["cards_inserted"] = 0
+            job.payload = payload
+            job.error_code = "all_dropped"
+            job.status = "completed"
+            job.completed_at = now()
+            return
+
+        # 5. 성공 — 구버전 카드 retire 후 append (dedup partial unique 충돌 회피).
+        await supersede_old_cards(
+            session, source_question_id=question.id, keep_generated_at=source_version
+        )
+        model_name = f"mlx:{primary_name}"
+        inserted = 0
+        for g in guarded:
+            card_id = await append_card(
+                session,
+                user_id=question.user_id,
+                study_topic_id=question.study_topic_id,
+                source_kind=SOURCE_KIND_QUESTION,
+                source_question_id=question.id,
+                format=g.format,
+                cue=g.cue,
+                fact=g.fact,
+                cloze_text=g.cloze_text,
+                dedup_hash=g.dedup_hash,
+                source_generated_at=source_version,
+                model=model_name,
+                generated_at=now(),
+                needs_review=True,
+            )
+            if card_id is not None:
+                inserted += 1
+                if g.matched_evidence:
+                    await append_card_evidence(
+                        session, card_id=card_id, refs=g.matched_evidence
+                    )
+
+        payload["cards_inserted"] = inserted
+        job.payload = payload
+        job.status = "completed"
+        job.completed_at = now()
+        return
+
+    except (asyncio.TimeoutError, httpx.HTTPError) as e:
+        job.error_code = "llm_timeout"
+        job.error_message = f"{type(e).__name__}: {e}"
+        logger.warning(
+            "card_extract_job_timeout job_id=%s src=%s/%s: %s",
+            job.id, job.source_kind, job.source_id, e,
+        )
+    except (json.JSONDecodeError, ValueError) as e:
+        job.error_code = "parse_fail"
+        job.error_message = f"{type(e).__name__}: {e}"
+        logger.warning(
+            "card_extract_job_parse_fail job_id=%s src=%s/%s: %s",
+            job.id, job.source_kind, job.source_id, e,
+        )
+    except Exception as e:
+        job.error_code = "unknown"
+        job.error_message = f"{type(e).__name__}: {e}"
+        logger.exception(
+            "card_extract_job_unknown_fail job_id=%s src=%s/%s",
+            job.id, job.source_kind, job.source_id,
+        )
+    finally:
+        # 재시도 분기 — all_dropped/no_ready/unsupported 는 위에서 이미 종결.
+        # 여기 도달 = llm_timeout / parse_fail / unknown.
+        if job.status == "processing":
+            retryable = job.error_code in ("llm_timeout", "parse_fail", "unknown")
+            if retryable and job.attempts < job.max_attempts:
+                job.status = "pending"
+            else:
+                job.status = "failed"
+                job.completed_at = now()
diff --git a/migrations/287_study_memo_cards.sql b/migrations/287_study_memo_cards.sql
new file mode 100644
index 0000000..6b5dcc5
--- /dev/null
+++ b/migrations/287_study_memo_cards.sql
@@ -0,0 +1,38 @@
+-- 287_study_memo_cards.sql
+-- 공부 암기노트 Phase 1: 추출 플래시카드 본체 (FK 트리 루트).
+--
+-- 출처(source_kind): question (P1 활성) / subject_note / document (P3 예약).
+-- 포맷(format): qa (cue->fact) / cloze (빈칸). 강한 enum 미사용 — read-time 매핑.
+-- needs_review DEFAULT true: 생성물이라 추출 직후 검토 대기 (study_questions 의 false 와 반대).
+-- source_generated_at: 추출 당시 study_questions.ai_explanation_generated_at — stale 판정/버전 핀.
+-- source_question_id 만 nullable + ON DELETE CASCADE (문제 물리삭제 시 카드 동반삭제;
+--   단 study_questions 는 soft-delete 만이라 실전은 정정/삭제 훅이 needs_review 마킹).
+-- 인용(evidence) 은 별 테이블 study_memo_card_evidence (append-only).
+
+CREATE TABLE IF NOT EXISTS study_memo_cards (
+    id                      BIGSERIAL    PRIMARY KEY,
+    user_id                 BIGINT       NOT NULL REFERENCES users(id)        ON DELETE CASCADE,
+    study_topic_id          BIGINT       NOT NULL REFERENCES study_topics(id) ON DELETE CASCADE,
+
+    source_kind             VARCHAR(40)  NOT NULL,
+    source_question_id      BIGINT       REFERENCES study_questions(id) ON DELETE CASCADE,
+    source_subject_note_id  BIGINT,
+
+    format                  VARCHAR(20)  NOT NULL,
+    cue                     TEXT         NOT NULL,
+    fact                    TEXT         NOT NULL,
+    cloze_text              TEXT,
+    extra                   JSONB,
+
+    source_generated_at     TIMESTAMPTZ,
+    dedup_hash              VARCHAR(64)  NOT NULL,
+
+    needs_review            BOOLEAN      NOT NULL DEFAULT true,
+    flagged_at              TIMESTAMPTZ,
+    flagged_by              VARCHAR(40),
+
+    model                   VARCHAR(120),
+    generated_at            TIMESTAMPTZ,
+    created_at              TIMESTAMPTZ  NOT NULL DEFAULT now(),
+    deleted_at              TIMESTAMPTZ
+);
diff --git a/migrations/288_study_memo_cards_dedup_uq.sql b/migrations/288_study_memo_cards_dedup_uq.sql
new file mode 100644
index 0000000..1c24285
--- /dev/null
+++ b/migrations/288_study_memo_cards_dedup_uq.sql
@@ -0,0 +1,10 @@
+-- 288_study_memo_cards_dedup_uq.sql
+-- dedup_hash 중복 카드 차단의 최종 방어선 (구조로 강제).
+-- append_card 의 ON CONFLICT (dedup_hash) DO NOTHING 이 매칭할 UNIQUE 제약 — 필수.
+-- PARTIAL (WHERE deleted_at IS NULL): supersede 로 retire 된 구버전 카드가
+--   같은 dedup_hash 의 새 추출을 막지 않도록 살아있는 카드만 유일성 강제.
+-- dedup_hash = sha256(source_question_id | format | normalize(정답토큰)).
+
+CREATE UNIQUE INDEX IF NOT EXISTS uq_study_memo_cards_dedup
+    ON study_memo_cards (dedup_hash)
+    WHERE deleted_at IS NULL;
diff --git a/migrations/289_study_memo_cards_source_q_idx.sql b/migrations/289_study_memo_cards_source_q_idx.sql
new file mode 100644
index 0000000..2047c0f
--- /dev/null
+++ b/migrations/289_study_memo_cards_source_q_idx.sql
@@ -0,0 +1,8 @@
+-- 289_study_memo_cards_source_q_idx.sql
+-- 정정/삭제 훅의 일괄 UPDATE (WHERE source_question_id=...) 와
+-- 워커 supersede (구버전 카드 retire) 조회 가속.
+-- PARTIAL (WHERE deleted_at IS NULL): 살아있는 카드만 색인.
+
+CREATE INDEX IF NOT EXISTS idx_study_memo_cards_source_q
+    ON study_memo_cards (source_question_id)
+    WHERE deleted_at IS NULL;
diff --git a/migrations/290_study_memo_card_evidence.sql b/migrations/290_study_memo_card_evidence.sql
new file mode 100644
index 0000000..8594ccc
--- /dev/null
+++ b/migrations/290_study_memo_card_evidence.sql
@@ -0,0 +1,15 @@
+-- 290_study_memo_card_evidence.sql
+-- 카드별 인용(citation) append-only 원장.
+-- card-context 가 모은 evidence_refs (source_type document|question, source_id, snippet)
+--   를 카드 추출 워커가 그대로 적재. UPDATE/DELETE 없음 — updated_at/deleted_at 미포함.
+-- card_id ON DELETE CASCADE: 카드 삭제 시 인용 동반삭제.
+
+CREATE TABLE IF NOT EXISTS study_memo_card_evidence (
+    id            BIGSERIAL    PRIMARY KEY,
+    card_id       BIGINT       NOT NULL REFERENCES study_memo_cards(id) ON DELETE CASCADE,
+    source_type   VARCHAR(40)  NOT NULL,
+    source_id     BIGINT,
+    chunk_index   INTEGER,
+    snippet       TEXT,
+    created_at    TIMESTAMPTZ  NOT NULL DEFAULT now()
+);
diff --git a/migrations/291_study_memo_card_jobs.sql b/migrations/291_study_memo_card_jobs.sql
new file mode 100644
index 0000000..3e56f54
--- /dev/null
+++ b/migrations/291_study_memo_card_jobs.sql
@@ -0,0 +1,29 @@
+-- 291_study_memo_card_jobs.sql
+-- card_extract 비동기 작업 큐 (231_study_question_jobs.sql 복제 + 다형 소스).
+--   라이프사이클: pending -> processing -> completed | failed | skipped
+--   error_code 권장값: parse_fail / llm_timeout / unknown (재시도 대상),
+--                      all_dropped (0장 생성, completed 로 종결해 재추출 차단),
+--                      no_ready_explanation (skipped, 비재시도).
+-- source_question_id 직접 FK 대신 source_kind/source_id 다형 참조 (question|subject_note|document).
+-- source_version = 추출 대상 study_questions.ai_explanation_generated_at (버전 멱등키) —
+--   폴러의 NOT EXISTS(... AND source_version=현재버전) 가 같은 버전 재추출을 차단.
+
+CREATE TABLE IF NOT EXISTS study_memo_card_jobs (
+    id              BIGSERIAL    PRIMARY KEY,
+    user_id         BIGINT       NOT NULL REFERENCES users(id) ON DELETE CASCADE,
+
+    source_kind     VARCHAR(40)  NOT NULL,
+    source_id       BIGINT       NOT NULL,
+    source_version  TIMESTAMPTZ,
+
+    kind            VARCHAR(40)  NOT NULL,
+    status          VARCHAR(20)  NOT NULL DEFAULT 'pending',
+    attempts        SMALLINT     NOT NULL DEFAULT 0,
+    max_attempts    SMALLINT     NOT NULL DEFAULT 2,
+    error_code      VARCHAR(40),
+    error_message   TEXT,
+    payload         JSONB,
+    created_at      TIMESTAMPTZ  NOT NULL DEFAULT now(),
+    started_at      TIMESTAMPTZ,
+    completed_at    TIMESTAMPTZ
+);
diff --git a/migrations/292_study_memo_card_jobs_active_uq.sql b/migrations/292_study_memo_card_jobs_active_uq.sql
new file mode 100644
index 0000000..3b0ab1f
--- /dev/null
+++ b/migrations/292_study_memo_card_jobs_active_uq.sql
@@ -0,0 +1,9 @@
+-- 292_study_memo_card_jobs_active_uq.sql
+-- (source_kind, source_id) 활성 행 중복 차단 (232 패턴).
+-- terminal status (completed/failed/skipped) 는 누적 이력이라 unique 대상 X.
+-- 동시 active 1행만 보장; 버전 멱등(같은 source_version 재추출 차단)은 폴러 NOT EXISTS 책임.
+-- 키에 source_version 을 넣지 않음 — 같은 (kind,id) 의 동시 active 추출은 1건이어야 함.
+
+CREATE UNIQUE INDEX IF NOT EXISTS uq_study_memo_card_jobs_active
+    ON study_memo_card_jobs (source_kind, source_id)
+    WHERE status IN ('pending', 'processing');
diff --git a/migrations/293_study_memo_card_jobs_lookup_idx.sql b/migrations/293_study_memo_card_jobs_lookup_idx.sql
new file mode 100644
index 0000000..7ec2307
--- /dev/null
+++ b/migrations/293_study_memo_card_jobs_lookup_idx.sql
@@ -0,0 +1,9 @@
+-- 293_study_memo_card_jobs_lookup_idx.sql
+-- 폴러(study_card_enqueue)의 버전 멱등 NOT EXISTS 조회 가속:
+--   NOT EXISTS (SELECT 1 FROM study_memo_card_jobs
+--               WHERE source_kind='question' AND source_id=sq.id
+--                 AND source_version=sq.ai_explanation_generated_at)
+-- terminal 행까지 전부 봐야 하므로 partial 아님(active uq 와 별개).
+
+CREATE INDEX IF NOT EXISTS idx_study_memo_card_jobs_lookup
+    ON study_memo_card_jobs (source_kind, source_id, source_version);
diff --git a/migrations/294_study_memo_card_progress.sql b/migrations/294_study_memo_card_progress.sql
new file mode 100644
index 0000000..d5b642d
--- /dev/null
+++ b/migrations/294_study_memo_card_progress.sql
@@ -0,0 +1,23 @@
+-- 294_study_memo_card_progress.sql
+-- 카드 SR(간격반복) 미러 — P1 휴면.
+-- P1 에서는 writer 가 없어 빈 테이블만 생성한다 (SR 산술/sr_schedule 공용추출은 P3).
+--   미리 만드는 이유: P3 에서 ALTER 없이 데이터 채우기만 하도록 스키마 선확보.
+-- 226_study_question_progress.sql 골격을 카드용으로 미러 (question -> card).
+-- 간격 상수 정본(P3 적용): REVIEW_INTERVAL_DAYS={1:3,2:7,3:14}, MASTERED=4, FIRST_DUE=1.
+
+CREATE TABLE IF NOT EXISTS study_memo_card_progress (
+    id                BIGSERIAL    PRIMARY KEY,
+    user_id           BIGINT       NOT NULL REFERENCES users(id)        ON DELETE CASCADE,
+    study_topic_id    BIGINT       NOT NULL REFERENCES study_topics(id) ON DELETE CASCADE,
+    card_id           BIGINT       NOT NULL REFERENCES study_memo_cards(id) ON DELETE CASCADE,
+
+    last_outcome      VARCHAR(20),
+    last_reviewed_at  TIMESTAMPTZ,
+    due_at            TIMESTAMPTZ,
+    review_stage      SMALLINT,
+
+    created_at        TIMESTAMPTZ  NOT NULL DEFAULT now(),
+    updated_at        TIMESTAMPTZ  NOT NULL DEFAULT now(),
+
+    CONSTRAINT uq_card_progress_user_card UNIQUE (user_id, card_id)
+);
diff --git a/migrations/295_study_topics_focused_at.sql b/migrations/295_study_topics_focused_at.sql
new file mode 100644
index 0000000..865dc10
--- /dev/null
+++ b/migrations/295_study_topics_focused_at.sql
@@ -0,0 +1,7 @@
+-- 295_study_topics_focused_at.sql
+-- 공부중 태그. focused_at IS NOT NULL = 포커스 중 (알람/세션-prep 대상).
+-- PATCH 토글로 set/clear. 폴러 초기 스코프 + reminder 스코프 술어가 참조.
+-- DEFAULT 없음 (NULL = 비포커스) — DEFAULT now() 면 기존 전 토픽이 포커스로 오인됨.
+
+ALTER TABLE study_topics
+    ADD COLUMN IF NOT EXISTS focused_at TIMESTAMPTZ;
diff --git a/migrations/296_study_questions_needs_review_cols.sql b/migrations/296_study_questions_needs_review_cols.sql
new file mode 100644
index 0000000..4eb41c5
--- /dev/null
+++ b/migrations/296_study_questions_needs_review_cols.sql
@@ -0,0 +1,11 @@
+-- 296_study_questions_needs_review_cols.sql
+-- 검토 대기 플래그 3컬럼 (정정/삭제 훅 + needs_review 큐가 set/clear).
+-- needs_review DEFAULT false: 기존 문제는 기본 정상 (study_memo_cards 의 true 와 반대).
+-- flagged_by 값은 서버측 상수만 적재: 'user' / 'source_changed' / 'source_deleted'
+--   (raw 사용자 입력 금지). 강한 enum 미사용 — read-time 매핑.
+-- NOT NULL DEFAULT false 는 PG11+ 메타데이터 fast-path (즉시) — 빈 시간대 배포 권장.
+
+ALTER TABLE study_questions
+    ADD COLUMN IF NOT EXISTS needs_review BOOLEAN NOT NULL DEFAULT false,
+    ADD COLUMN IF NOT EXISTS flagged_at   TIMESTAMPTZ,
+    ADD COLUMN IF NOT EXISTS flagged_by   VARCHAR(40);
diff --git a/migrations/297_study_questions_needs_review_idx.sql b/migrations/297_study_questions_needs_review_idx.sql
new file mode 100644
index 0000000..a6105c6
--- /dev/null
+++ b/migrations/297_study_questions_needs_review_idx.sql
@@ -0,0 +1,9 @@
+-- 297_study_questions_needs_review_idx.sql
+-- needs_review 큐 뷰(GET /study-questions?needs_review=true) + count 용 부분 인덱스.
+-- WHERE 술어(deleted_at IS NULL AND needs_review)는 큐 뷰 쿼리 WHERE 와 글자 단위로
+--   일치해야 partial index 가 선택된다 (HR-5 쿼리와 정합 필수).
+-- soft-delete 행 제외(deleted_at IS NULL 합류).
+
+CREATE INDEX IF NOT EXISTS idx_study_questions_needs_review
+    ON study_questions (study_topic_id)
+    WHERE deleted_at IS NULL AND needs_review;
diff --git a/migrations/298_study_reminders.sql b/migrations/298_study_reminders.sql
new file mode 100644
index 0000000..268ad0e
--- /dev/null
+++ b/migrations/298_study_reminders.sql
@@ -0,0 +1,18 @@
+-- 298_study_reminders.sql
+-- 알람 재료 append-only 원장. study_reminder cron(09/13/19 KST)이 발화 시 1행 INSERT,
+--   GET /reminders/latest 가 읽는다. UPDATE/DELETE 없음.
+-- fired_at 은 발화 시각의 '시간 슬롯' 으로 truncate 해서 박는다 (raw now() 마이크로초면
+--   UNIQUE 가 사실상 안 걸려 멱등 무효). UNIQUE(user_id, fired_at) + on_conflict_do_nothing.
+-- study_topic_id ON DELETE SET NULL: 토픽 삭제돼도 과거 알람 이력 보존(CASCADE 아님).
+
+CREATE TABLE IF NOT EXISTS study_reminders (
+    id                 BIGSERIAL    PRIMARY KEY,
+    user_id            BIGINT       NOT NULL REFERENCES users(id)        ON DELETE CASCADE,
+    study_topic_id     BIGINT       REFERENCES study_topics(id)          ON DELETE SET NULL,
+    due_count          INTEGER,
+    focus_topic_names  JSONB,
+    fired_at           TIMESTAMPTZ  NOT NULL,
+    created_at         TIMESTAMPTZ  NOT NULL DEFAULT now(),
+
+    CONSTRAINT uq_study_reminders_user_fired UNIQUE (user_id, fired_at)
+);