From b0894f000dc262028665380f902a354d7c034ea9 Mon Sep 17 00:00:00 2001
From: "sungkuk.chae-at-471112812288"
 <76669236+cskwork@users.noreply.github.com>
Date: Wed, 23 Sep 2026 09:41:27 +0900
Subject: [PATCH 1/4] feat: area pages record how a figure is calculated
 (v0.16.0)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

- templates/area.md gains an optional Numbers (통계 산정) section between
  Business rules and How it works: one line per on-screen count, rate,
  score or chart figure, numbered N1... with the same permanence rules as
  a business rule (P-number)
- Extended only where omitting N-numbers would leave figures unprotected:
  templates/spec.md's Business rules touched section, ship's retrospective
  area-candidate instruction, templates/harvest.md's Area candidates
  format, and AGENTS.md rule 4's knowledge-filing table
- README.md / README.ko.md: one mention each in the product-area paragraph
  and the memory-tree comment
- kb.sh untouched: the Rules count stays P-only, no correctness gap

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
---
 AGENTS.md                |  1 +
 README.ko.md             |  4 ++--
 README.md                |  4 ++--
 VERSION                  |  2 +-
 log/changelog-v0.16.0.md | 51 ++++++++++++++++++++++++++++++++++++++++
 skills/5-ship/SKILL.md   |  4 +++-
 templates/area.md        |  7 ++++++
 templates/harvest.md     |  6 +++--
 templates/spec.md        |  1 +
 9 files changed, 72 insertions(+), 8 deletions(-)
 create mode 100644 log/changelog-v0.16.0.md

diff --git a/AGENTS.md b/AGENTS.md
index 428ba86..8409a28 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -148,6 +148,7 @@ archives the feature to `.sdlc/archive/<slug>/`.
    | What | Where | Written by |
    |---|---|---|
    | A product's business rule (정책) — "a submitted answer cannot be edited" | `memory/areas/<area-slug>.md` Business rules, numbered P1… (templates/area.md) | close merge, shipped only |
+   | How an on-screen count, rate, score or chart figure is calculated | the same page's Numbers, numbered N1… (templates/area.md) | close merge, shipped only |
    | What changed in an area, when | the area page's History, one line per feature | close merge, shipped only |
    | A fact that holds for one area only | the area page's How it works | close merge |
    | A term, or a system fact/constraint that spans areas | `memory/DOMAIN.md` (100 lines max) | close merge |
diff --git a/README.ko.md b/README.ko.md
index 5b52b32..5758fa6 100644
--- a/README.ko.md
+++ b/README.ko.md
@@ -145,7 +145,7 @@ agent  APPROVED: intent of claims-status (.sdlc/work/claims-status/intent.md)
 │   ├── POLICY.md                     # 사람이 선언한 하드 룰, 에이전트는 전사만
 │   ├── INDEX.md                      # 교훈 포인터, 50줄 이하
 │   ├── DOMAIN.md                     # 여러 제품 영역에 걸친 용어 · 사실 · 제약
-│   ├── areas/<area-slug>.md               # 제품 영역(웹앱은 메뉴)마다 한 장: 업무 정책 P1… · 동작 · 변경 이력
+│   ├── areas/<area-slug>.md               # 제품 영역(웹앱은 메뉴)마다 한 장: 업무 정책 P1… · 통계 산정 N1…(선택) · 동작 · 변경 이력
 │   └── lessons/<date>-<lesson>.md
 ├── work/<slug>/                      # 열린 피처만
 │   ├── origin.md                     # 요청 당시의 티켓 · 기획서 스냅샷 — intent 게이트가 결합
@@ -169,7 +169,7 @@ agent  APPROVED: intent of claims-status (.sdlc/work/claims-status/intent.md)
 
 **기록을 어디에 둘지는 사용자가 정합니다.** 기본값은 프로젝트 작업 사본 안이고, `init.sh . --area ~/knowledge`를 쓰면 사용자가 고른 폴더 아래 `<area>/<단위이름>-<체크아웃 식별자>/`에 저장하고 `.sdlc`를 그곳으로 연결합니다. 체크아웃마다 저장소가 하나씩이므로 워크트리 두 개가 승인 상태를 공유하는 일이 없습니다. 영역이 프로젝트 안에 있거나 프로젝트가 영역 안에 있을 때, 다른 체크아웃이 이미 그 저장소를 소유할 때, 실제 `.sdlc` 디렉터리가 이미 있을 때(자동으로 옮기지 않습니다), 링크를 만들 수 없을 때는 아무것도 쓰지 않고 분명히 실패합니다. 이 소유권은 init 시점뿐 아니라 실행 시점에도 다시 확인합니다. `<store>/PROJECT`에 적힌 체크아웃이 지금 실행 중인 체크아웃과 다르면 `check-gate.sh`, `approve.sh`, `close.sh`, `status.sh`, `tools/auto.sh`, `tools/verify.sh`, `tools/handoff.sh`가 판정을 내리거나 상태를 쓰기 전에 거부하므로, 심볼릭 링크를 그대로 복사한 작업 사본(`cp -R`, rsync, 대부분의 백업 복원)이 다른 체크아웃의 게이트를 열거나 그 피처를 닫을 수 없습니다. 읽기는 이 제약을 받지 않아 `tools/kb.sh show|search|list`는 그대로 쓸 수 있고, 소유권을 자동으로 옮기거나 다시 묶는 일은 없습니다. 어느 쪽을 고르든 **저장소 백업은 사용자의 몫입니다.** git이 더 이상 대신해 주지 않습니다.
 
-**지식은 제품 영역별로 정리됩니다.** 웹앱이면 제품 영역은 메뉴 하나이고 메뉴 경로(`학습 > 평가 > 제출`)로 부릅니다. 다른 소프트웨어는 모듈, API, 배치 작업, CLI 명령입니다. 제품 영역마다 `memory/areas/<area-slug>.md` 한 장이 있고, 여기에 개발자가 아니어도 읽을 수 있는 문장으로 쓴 업무 정책(P1, P2…), 동작 방식, 그 영역을 바꾼 피처별 이력 한 줄이 담깁니다. 피처의 `summary.md`는 `Area:` 줄로 제품 영역을 적고, spec은 어떤 정책을 유지하거나 바꾸는지 밝히고, Side effects 검증자는 건드리지 않아야 할 정책이 그대로인지 다시 확인하며, close 병합이 새로 생기거나 바뀐 정책을 페이지에 올립니다. `tools/kb.sh show "학습 > 평가 > 제출"`(또는 페이지 파일 이름)은 그 페이지와 그 제품 영역을 바꾼 피처를 함께 보여줍니다.
+**지식은 제품 영역별로 정리됩니다.** 웹앱이면 제품 영역은 메뉴 하나이고 메뉴 경로(`학습 > 평가 > 제출`)로 부릅니다. 다른 소프트웨어는 모듈, API, 배치 작업, CLI 명령입니다. 제품 영역마다 `memory/areas/<area-slug>.md` 한 장이 있고, 여기에 개발자가 아니어도 읽을 수 있는 문장으로 쓴 업무 정책(P1, P2…), 동작 방식, 그 영역을 바꾼 피처별 이력 한 줄이 담깁니다. 건수·비율·점수·차트를 보여주는 영역은 통계 산정(N1, N2…) 섹션도 가지며, 화면의 수치 하나마다 무엇을 세는지와 출처를 한 줄로 남깁니다 — 업무 정책과 같은 영속성·처리 방식을 따릅니다. 피처의 `summary.md`는 `Area:` 줄로 제품 영역을 적고, spec은 어떤 정책이나 수치를 유지하거나 바꾸는지 밝히고, Side effects 검증자는 건드리지 않아야 할 것이 그대로인지 다시 확인하며, close 병합이 새로 생기거나 바뀐 정책·수치를 페이지에 올립니다. `tools/kb.sh show "학습 > 평가 > 제출"`(또는 페이지 파일 이름)은 그 페이지와 그 제품 영역을 바꾼 피처를 함께 보여줍니다.
 
 **기록을 다시 읽는 도구는 `tools/kb.sh`입니다.** `index`는 목차 페이지를 다시 만듭니다(`init.sh`와 `close.sh`가 자동으로 실행합니다). 페이지는 정책 수·마지막 변경·피처를 담은 제품 영역 표, 상태·날짜·제품 영역·태그를 담은 개요 표(최신순), 아직 close가 병합하지 않은 harvest 목록, 피처별 절 순서입니다. `show <slug>`는 피처 하나를 요약본으로 보여줍니다. 목표, `summary.md`(제품 영역·무엇이 문제였나·Before → After·확인 방법을 담는, 계속 갱신하도록 만든 유일한 기록), 배포 상태, 병합되지 않은 harvest 후보, 교훈 제목이 먼저 나오고 파일 경로는 마지막입니다. `search "<문자열>"`은 열린 피처와 닫힌 피처, 지속 메모리를 대상으로 출력량을 제한한 문자열 검색을 하고, `harvest [--stale <일수>]`는 harvest.md가 아직 memory/에 들어가지 않은 열린 피처를 유휴 기간과 함께 나열합니다(유휴 상태가 오래된 피처는 close 없이 병합할 수 있습니다 — AGENTS.md 규칙 4). `--area <폴더>`를 붙이면 그 폴더 안의 모든 저장소를 대상으로 같은 일을 하며, 원래 체크아웃이 사라진 피처도 읽을 수 있습니다. 저장소 config.md에 `index_style: obsidian`을 적으면 Obsidian 볼트용 frontmatter와 인라인 `#태그`를 덧붙입니다. 생성 시각은 절대 쓰지 않으므로 내용이 같으면 diff도 생기지 않습니다. 종료 코드는 `0` 찾음, `1` 없음, `2` 사용법 오류 또는 거부입니다.
 
diff --git a/README.md b/README.md
index f025bd4..2d56793 100644
--- a/README.md
+++ b/README.md
@@ -145,7 +145,7 @@ Per feature, inside the **target project**:
 │   ├── POLICY.md                     # human-declared hard rules; agents transcribe only
 │   ├── INDEX.md                      # ≤50 lines of lesson pointers
 │   ├── DOMAIN.md                     # terms · facts and constraints that span areas
-│   ├── areas/<area-slug>.md               # one per product area (web app: one menu): business rules P1… · how it works · history
+│   ├── areas/<area-slug>.md               # one per product area (web app: one menu): business rules P1… · numbers N1… (optional) · how it works · history
 │   └── lessons/<date>-<lesson>.md
 ├── work/<slug>/                      # OPEN features only
 │   ├── origin.md                     # the ticket / 기획서 as requested — bound by the intent gate
@@ -169,7 +169,7 @@ Per feature, inside the **target project**:
 
 **Where the records live is your choice.** By default they sit in the project's working copy. `init.sh . --area ~/knowledge` puts them in a folder you choose instead — `<area>/<unit>-<checkout-id>/`, with `.sdlc` linked to it, one store per checkout so two worktrees never share approvals. The area is refused if it sits inside the project (or the project inside it), if another checkout already owns that store, if a real `.sdlc` directory is already there (nothing is ever relocated for you), or if the link cannot be made. That ownership is re-checked at RUNTIME, not only at init: `check-gate.sh`, `approve.sh`, `close.sh`, `status.sh`, `tools/auto.sh`, `tools/verify.sh` and `tools/handoff.sh` refuse before any verdict or write when `<store>/PROJECT` names a different checkout, so a copied working copy (`cp -R`, rsync and most restores keep the symlink) can neither open another checkout's gate nor close its features. Reading is never bound that way: `tools/kb.sh show|search|list` still works, and nothing is ever re-bound or moved for you. Whichever you choose, **the store is yours to back up** — git no longer does it for you.
 
-**Knowledge is filed by product area.** For a web app an area is one menu, named by its menu path (`학습 > 평가 > 제출`); for other software a module, API, job, or CLI command. Each area has one page, `memory/areas/<area-slug>.md`: its business rules numbered P1, P2… in sentences a non-developer can read, how it works, and a history line per feature that changed it. A feature's `summary.md` names its area on an `Area:` line, the spec states which rules it keeps or changes, the Side effects verifier re-checks the rules it should not have touched, and the close merge files new or changed rules on the page. `tools/kb.sh show "학습 > 평가 > 제출"` (or the page's file name) prints the page with the features that changed it.
+**Knowledge is filed by product area.** For a web app an area is one menu, named by its menu path (`학습 > 평가 > 제출`); for other software a module, API, job, or CLI command. Each area has one page, `memory/areas/<area-slug>.md`: its business rules numbered P1, P2… in sentences a non-developer can read, how it works, and a history line per feature that changed it. An area that shows counts, rates, scores or charts also gets a Numbers section, N1, N2…, one line per figure recording what it counts and its source — the same permanence and treatment as a business rule. A feature's `summary.md` names its area on an `Area:` line, the spec states which rules or figures it keeps or changes, the Side effects verifier re-checks the ones it should not have touched, and the close merge files new or changed rules and figures on the page. `tools/kb.sh show "학습 > 평가 > 제출"` (or the page's file name) prints the page with the features that changed it.
 
 **Reading the records back** is `tools/kb.sh`: `index` regenerates the contents page (`init.sh` and `close.sh` do it for you) — the product areas with their rule count, last change and features, an overview table by state, date, area and tags, newest first, the harvests no close has merged yet, then one section per feature; `show <slug>` prints one feature as a digest — goal, its `summary.md` (area, what was wrong, before → after, how to check, the one record meant to be kept current), delivery, unmerged harvest candidates, lesson titles, then the paths; `search "<text>"` does a bounded literal search over open and closed features plus durable memory; `harvest [--stale <days>]` lists open features whose harvest.md is not in memory yet, with idle time (a stale one may be merged without closing — AGENTS.md rule 4); and `--area <folder>` does any of these across every store in that folder — including features whose checkout no longer exists. `index_style: obsidian` in the store's config.md adds frontmatter and inline `#tags` for a vault; no timestamp is ever written, so an unchanged page produces no diff. Exit codes: `0` found, `1` nothing found, `2` usage error or refusal.
 
diff --git a/VERSION b/VERSION
index 86dd09a..cfe6c00 100644
--- a/VERSION
+++ b/VERSION
@@ -1 +1 @@
-v0.15.0
+v0.16.0
diff --git a/log/changelog-v0.16.0.md b/log/changelog-v0.16.0.md
new file mode 100644
index 0000000..f0223f1
--- /dev/null
+++ b/log/changelog-v0.16.0.md
@@ -0,0 +1,51 @@
+# v0.16.0 — area pages record how a figure is calculated
+
+v0.15.0 gave a product area one home for its business rules (P-numbers). A
+menu that shows counts, rates, scores or charts had nowhere to say what a
+number on screen actually means — "what is counted, out of what, real-time
+or batch" lived in someone's head or nowhere. This release adds an optional
+Numbers section to the area page, numbered N1… with the same permanence and
+filing rules as a business rule. Gates, approvals, the `--json` schema, and
+`kb.sh`'s rule count are unchanged.
+
+## Changes
+
+- **`templates/area.md`** gains an optional **Numbers (통계 산정)** section
+  between Business rules and How it works: one line per on-screen figure,
+  named as the screen labels it, recording what it counts, out of what,
+  whether it is real-time or batch (job, refresh, data-as-of), its source,
+  and who set it. Delete the section for an area with no figures. N-numbers
+  follow the P-number rules: permanent, a changed figure is edited in place,
+  a retired one stays struck through.
+- **The loop treats a figure like a business rule** wherever omitting it
+  would leave it unprotected:
+  - `templates/spec.md`'s "Business rules touched" section gains a sibling
+    line for numbers touched (kept/changed/new), so the Side effects
+    verifier re-checks a figure it did not set out to change, the same way
+    it already re-checks a P-number.
+  - Ship's retrospective (`skills/5-ship/SKILL.md`) writes a figure a
+    feature sets, changes, or retires as an area candidate, same as a
+    business rule.
+  - `templates/harvest.md`'s Area candidates format gains the `N<n>`
+    line shape the close merge numbers, next to the existing `P<n>` one.
+  - AGENTS.md rule 4's "what goes where" table gains a row: a figure's
+    calculation method files to the area page's Numbers, written by the
+    close merge on a shipped close only — same as a business rule.
+- **README.md / README.ko.md**: the "Knowledge is filed by product area"
+  paragraph and the memory-tree comment now mention the optional Numbers
+  section.
+
+## Compatibility
+
+An area page with no Numbers section is unaffected: the section is
+optional and `kb.sh`'s Rules column stays a P-number count only (a store
+with only P-numbers renders identically to before). No script parses the
+`N<n>:` line; filing it is an instruction to the agent, exactly like a
+`P<n>:` line.
+
+## Validation
+
+- `bash gates/selftest.sh` → `SELFTEST PASS` (unchanged; the smoke test
+  does not exercise the Numbers section, and this change does not touch
+  the scripts it does cover — gate mechanics, lazymode, close proof,
+  product-area filing by menu path).
diff --git a/skills/5-ship/SKILL.md b/skills/5-ship/SKILL.md
index 0fb0661..f036324 100644
--- a/skills/5-ship/SKILL.md
+++ b/skills/5-ship/SKILL.md
@@ -69,7 +69,9 @@ AGENTS.md rule 4). **Every business rule this feature set, changed, or
 retired is an area candidate** — the rule in one testable sentence, its
 source (the origin, the spec R-item or compact intent O-item), the P-number
 it changes or retires — plus one
-history line per area it changed. Then finish the feature's `summary.md`:
+history line per area it changed. **A figure the area page shows a Numbers
+section for is treated the same way**: the N-number it sets, changes, or
+retires, named on screen, with what it counts and its source. Then finish the feature's `summary.md`:
 `Status` says only what delivery.md confirms (a pushed review branch is not
 a deployment), Before → After and How to check match what was actually
 built and proven, Remember holds the one thing worth knowing next time.
diff --git a/templates/area.md b/templates/area.md
index a03db05..2f7e3f3 100644
--- a/templates/area.md
+++ b/templates/area.md
@@ -22,6 +22,13 @@
        - ~~P3: <rule>~~ — retired YYYY-MM-DD by <slug>: <why>
      Not POLICY.md — that file holds the human's rules for agents. -->
 
+## Numbers (통계 산정)
+<!-- Only for an area that shows counts, rates, scores or charts; delete the
+     section otherwise. One line per figure, named as the screen labels it:
+       - N1: <label on screen> — <what is counted> / <out of what> · <real-time | batch: <job>, refreshed <when>, data up to <when>> — source: <기획서 / ticket / code, date> · set by <slug> · [verified: how — YYYY-MM-DD]
+     N-numbers follow the P-number rules above: permanent, a changed figure is
+     edited in place, a retired one is kept struck through. -->
+
 ## How it works
 <!-- ≤10 lines: what the user does, step by step, what the system does in
      response, and facts that hold for this area only. Use DOMAIN.md terms;
diff --git a/templates/harvest.md b/templates/harvest.md
index 6245a32..4c43802 100644
--- a/templates/harvest.md
+++ b/templates/harvest.md
@@ -7,10 +7,12 @@
 
 ## Area candidates
 <!-- - [area: <menu path>] rule new | P<n> changed | P<n> retired: <rule> — source · [verified: how — YYYY-MM-DD]
+     - [area: <menu path>] figure new | N<n> changed | N<n> retired: <label on screen> — what it counts · source · [verified: how — YYYY-MM-DD]
      - [area: <menu path>] fact: <holds for this area only> — [verified: how — YYYY-MM-DD]
      - [area: <menu path>] history: YYYY-MM-DD <slug> — <what changed for the user>
-     Rules and history merge only when the feature closes shipped; a stale
-     merge leaves them here (rule 4). The merge numbers a new rule. -->
+     Rules, figures, and history merge only when the feature closes shipped;
+     a stale merge leaves them here (rule 4). The merge numbers a new rule
+     or figure. -->
 
 ## Domain candidates
 <!-- - fact / term / constraint that spans areas — [verified: how — YYYY-MM-DD]
diff --git a/templates/spec.md b/templates/spec.md
index 2681b48..3b65df9 100644
--- a/templates/spec.md
+++ b/templates/spec.md
@@ -35,6 +35,7 @@ cases where they apply.>
 
 ## Business rules touched   <!-- from the area page(s) summary.md names; the verifier re-checks every kept one -->
 - <area> P1: kept · P3: changed → R2 · P4: retired → R5 · new rule → R6
+- <area> N1: kept · N2: changed → R2 · new figure → R6   <!-- only if the area page has a Numbers section; same treatment as business rules -->
 
 ## What stays untouched   <!-- brownfield: testable statements; becomes regression baseline -->
 - U1: <existing behavior>; checked by <command/test>

From 4ceb1871bf824ca8912ae8d3cadd8361c6566429 Mon Sep 17 00:00:00 2001
From: "sungkuk.chae-at-471112812288"
 <76669236+cskwork@users.noreply.github.com>
Date: Wed, 23 Sep 2026 12:56:51 +0900
Subject: [PATCH 2/4] feat: area pages named by menu path, laid out reader
 first (v0.16.0)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

- File name = menu path: the Menu line with each " > " written " - " and
  any of / \ : * ? " < > | replaced with "-" (교사 > 학생 > 학급 분석 →
  memory/areas/교사 - 학생 - 학급 분석.md). The ASCII kebab-case instruction
  is gone; AGENTS.md rule 4 and ship's retrospective tell the close merge
  the rule when it creates a page
- tools/kb.sh finds a page by file name, by Menu, or by the file name a
  Menu gives (kb_area_file); still refuses /, \ and a leading . as a file
  lookup; contents-page links percent-encode space % ( ) [ ] # < >
  (kb_area_href) so they open in GitHub and Obsidian; <summary> becomes a
  section label so show prints the plain part first, evidence last; an
  unfilled "- P1: <...>" template line is not counted or printed
- templates/area.md is reader first: plain rules, numbers, how it works and
  history on top; Where and one "P1 — source · set by · [verified]"
  evidence line per rule/figure inside a <details> block at the bottom; a
  retired rule's reason lives on its evidence line. Headings unchanged
- harvest.md, ship SKILL, AGENTS.md table: sources go to the evidence line
- selftest covers a Hangul menu-named page with spaces (lookup by Menu and
  file name, percent-encoded link, rule count ignores evidence/retired
  lines, path lookup refused)
- changelog v0.16.0 extended with a compatibility note: ASCII-named pages
  keep working; how to rename

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
---
 AGENTS.md                | 16 ++++++---
 README.ko.md             |  4 +--
 README.md                |  4 +--
 gates/selftest.sh        | 15 ++++++++
 init.sh                  |  2 +-
 log/changelog-v0.16.0.md | 74 +++++++++++++++++++++++++++++++++++-----
 skills/5-ship/SKILL.md   | 17 +++++----
 templates/area.md        | 68 +++++++++++++++++++++++-------------
 templates/harvest.md     |  8 +++--
 tools/kb.sh              | 57 +++++++++++++++++++++++--------
 10 files changed, 200 insertions(+), 65 deletions(-)

diff --git a/AGENTS.md b/AGENTS.md
index 8409a28..ae98475 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -147,8 +147,9 @@ archives the feature to `.sdlc/archive/<slug>/`.
 
    | What | Where | Written by |
    |---|---|---|
-   | A product's business rule (정책) — "a submitted answer cannot be edited" | `memory/areas/<area-slug>.md` Business rules, numbered P1… (templates/area.md) | close merge, shipped only |
+   | A product's business rule (정책) — "a submitted answer cannot be edited" | `memory/areas/<menu path>.md` Business rules, numbered P1…, one plain sentence each (templates/area.md) | close merge, shipped only |
    | How an on-screen count, rate, score or chart figure is calculated | the same page's Numbers, numbered N1… (templates/area.md) | close merge, shipped only |
+   | Where a rule or figure came from, and the code behind the area | the same page's evidence block at the bottom (`<details>`): `P1 — source · set by · [verified]`, the Where line | close merge |
    | What changed in an area, when | the area page's History, one line per feature | close merge, shipped only |
    | A fact that holds for one area only | the area page's How it works | close merge |
    | A term, or a system fact/constraint that spans areas | `memory/DOMAIN.md` (100 lines max) | close merge |
@@ -160,7 +161,12 @@ archives the feature to `.sdlc/archive/<slug>/`.
    named by its menu path (`학습 > 평가 > 제출`); otherwise a module, API,
    job, or CLI command — not the `--area` knowledge folder of rule 7.
    summary.md's `Area:` line and the page's `Menu:` line use the same words,
-   which is how `kb.sh` links them. DOMAIN over its limit: move
+   which is how `kb.sh` links them. The page's file name is that Menu with
+   each ` > ` written ` - ` and any of `/ \ : * ? " < > |` replaced with `-`
+   (`교사 > 학생 > 학급 분석` → `memory/areas/교사 - 학생 - 학급 분석.md`).
+   An area page reads reader first: rules, figures, how it works and history
+   in plain language on top; every source, verification, and code identifier
+   in the evidence block at the bottom. DOMAIN over its limit: move
    area-specific facts to their area page. INDEX over its limit: merge
    near-duplicates, drop superseded entries, replace promoted ones
    (skills/6-maintain).
@@ -168,8 +174,10 @@ archives the feature to `.sdlc/archive/<slug>/`.
    stages and researchers append candidates — one line each — to
    `.sdlc/work/<slug>/harvest.md` (templates/harvest.md). At close, merge
    every candidate into its row above — creating an area page from the
-   template when none exists, numbering new rules, adding the feature's
-   History line to each area it changed — then delete harvest.md. Business
+   template when none exists (file named by its menu path, as above),
+   numbering new rules, putting each rule's or figure's source on its
+   evidence line, adding the feature's History line to each area it
+   changed — then delete harvest.md. Business
    rules and History describe what the product DOES, so they merge only when
    the feature closes `shipped`: any other close drops them (summary.md keeps
    the story), and a stale merge leaves them in harvest.md for that close;
diff --git a/README.ko.md b/README.ko.md
index 5758fa6..3bad008 100644
--- a/README.ko.md
+++ b/README.ko.md
@@ -145,7 +145,7 @@ agent  APPROVED: intent of claims-status (.sdlc/work/claims-status/intent.md)
 │   ├── POLICY.md                     # 사람이 선언한 하드 룰, 에이전트는 전사만
 │   ├── INDEX.md                      # 교훈 포인터, 50줄 이하
 │   ├── DOMAIN.md                     # 여러 제품 영역에 걸친 용어 · 사실 · 제약
-│   ├── areas/<area-slug>.md               # 제품 영역(웹앱은 메뉴)마다 한 장: 업무 정책 P1… · 통계 산정 N1…(선택) · 동작 · 변경 이력
+│   ├── areas/<메뉴 경로>.md               # 제품 영역(웹앱은 메뉴)마다 한 장, 예: "학습 - 평가 - 제출.md": 업무 정책 P1… · 통계 산정 N1…(선택) · 동작 · 변경 이력 · 근거(접힘)
 │   └── lessons/<date>-<lesson>.md
 ├── work/<slug>/                      # 열린 피처만
 │   ├── origin.md                     # 요청 당시의 티켓 · 기획서 스냅샷 — intent 게이트가 결합
@@ -169,7 +169,7 @@ agent  APPROVED: intent of claims-status (.sdlc/work/claims-status/intent.md)
 
 **기록을 어디에 둘지는 사용자가 정합니다.** 기본값은 프로젝트 작업 사본 안이고, `init.sh . --area ~/knowledge`를 쓰면 사용자가 고른 폴더 아래 `<area>/<단위이름>-<체크아웃 식별자>/`에 저장하고 `.sdlc`를 그곳으로 연결합니다. 체크아웃마다 저장소가 하나씩이므로 워크트리 두 개가 승인 상태를 공유하는 일이 없습니다. 영역이 프로젝트 안에 있거나 프로젝트가 영역 안에 있을 때, 다른 체크아웃이 이미 그 저장소를 소유할 때, 실제 `.sdlc` 디렉터리가 이미 있을 때(자동으로 옮기지 않습니다), 링크를 만들 수 없을 때는 아무것도 쓰지 않고 분명히 실패합니다. 이 소유권은 init 시점뿐 아니라 실행 시점에도 다시 확인합니다. `<store>/PROJECT`에 적힌 체크아웃이 지금 실행 중인 체크아웃과 다르면 `check-gate.sh`, `approve.sh`, `close.sh`, `status.sh`, `tools/auto.sh`, `tools/verify.sh`, `tools/handoff.sh`가 판정을 내리거나 상태를 쓰기 전에 거부하므로, 심볼릭 링크를 그대로 복사한 작업 사본(`cp -R`, rsync, 대부분의 백업 복원)이 다른 체크아웃의 게이트를 열거나 그 피처를 닫을 수 없습니다. 읽기는 이 제약을 받지 않아 `tools/kb.sh show|search|list`는 그대로 쓸 수 있고, 소유권을 자동으로 옮기거나 다시 묶는 일은 없습니다. 어느 쪽을 고르든 **저장소 백업은 사용자의 몫입니다.** git이 더 이상 대신해 주지 않습니다.
 
-**지식은 제품 영역별로 정리됩니다.** 웹앱이면 제품 영역은 메뉴 하나이고 메뉴 경로(`학습 > 평가 > 제출`)로 부릅니다. 다른 소프트웨어는 모듈, API, 배치 작업, CLI 명령입니다. 제품 영역마다 `memory/areas/<area-slug>.md` 한 장이 있고, 여기에 개발자가 아니어도 읽을 수 있는 문장으로 쓴 업무 정책(P1, P2…), 동작 방식, 그 영역을 바꾼 피처별 이력 한 줄이 담깁니다. 건수·비율·점수·차트를 보여주는 영역은 통계 산정(N1, N2…) 섹션도 가지며, 화면의 수치 하나마다 무엇을 세는지와 출처를 한 줄로 남깁니다 — 업무 정책과 같은 영속성·처리 방식을 따릅니다. 피처의 `summary.md`는 `Area:` 줄로 제품 영역을 적고, spec은 어떤 정책이나 수치를 유지하거나 바꾸는지 밝히고, Side effects 검증자는 건드리지 않아야 할 것이 그대로인지 다시 확인하며, close 병합이 새로 생기거나 바뀐 정책·수치를 페이지에 올립니다. `tools/kb.sh show "학습 > 평가 > 제출"`(또는 페이지 파일 이름)은 그 페이지와 그 제품 영역을 바꾼 피처를 함께 보여줍니다.
+**지식은 제품 영역별로 정리됩니다.** 웹앱이면 제품 영역은 메뉴 하나이고 메뉴 경로(`학습 > 평가 > 제출`)로 부릅니다. 다른 소프트웨어는 모듈, API, 배치 작업, CLI 명령입니다. 제품 영역마다 페이지가 한 장 있고, 파일 이름은 메뉴 경로입니다 — `memory/areas/학습 - 평가 - 제출.md`(` > `는 ` - `로, `/ \ : * ? " < > |`는 `-`로 바꿉니다). 페이지는 읽는 사람 우선입니다. 위에는 개발자가 아니어도 읽을 수 있는 문장으로 쓴 업무 정책(P1, P2…), 동작 방식, 그 영역을 바꾼 피처별 이력 한 줄이 오고, 출처·검증·코드 위치는 모두 맨 아래 접힌 근거 블록에 둡니다. 건수·비율·점수·차트를 보여주는 영역은 통계 산정(N1, N2…) 섹션도 가지며, 화면의 수치 하나마다 무엇을 세는지와 출처를 한 줄로 남깁니다 — 업무 정책과 같은 영속성·처리 방식을 따릅니다. 피처의 `summary.md`는 `Area:` 줄로 제품 영역을 적고, spec은 어떤 정책이나 수치를 유지하거나 바꾸는지 밝히고, Side effects 검증자는 건드리지 않아야 할 것이 그대로인지 다시 확인하며, close 병합이 새로 생기거나 바뀐 정책·수치를 페이지에 올립니다. `tools/kb.sh show "학습 > 평가 > 제출"`(또는 페이지 파일 이름)은 그 페이지와 그 제품 영역을 바꾼 피처를 함께 보여줍니다.
 
 **기록을 다시 읽는 도구는 `tools/kb.sh`입니다.** `index`는 목차 페이지를 다시 만듭니다(`init.sh`와 `close.sh`가 자동으로 실행합니다). 페이지는 정책 수·마지막 변경·피처를 담은 제품 영역 표, 상태·날짜·제품 영역·태그를 담은 개요 표(최신순), 아직 close가 병합하지 않은 harvest 목록, 피처별 절 순서입니다. `show <slug>`는 피처 하나를 요약본으로 보여줍니다. 목표, `summary.md`(제품 영역·무엇이 문제였나·Before → After·확인 방법을 담는, 계속 갱신하도록 만든 유일한 기록), 배포 상태, 병합되지 않은 harvest 후보, 교훈 제목이 먼저 나오고 파일 경로는 마지막입니다. `search "<문자열>"`은 열린 피처와 닫힌 피처, 지속 메모리를 대상으로 출력량을 제한한 문자열 검색을 하고, `harvest [--stale <일수>]`는 harvest.md가 아직 memory/에 들어가지 않은 열린 피처를 유휴 기간과 함께 나열합니다(유휴 상태가 오래된 피처는 close 없이 병합할 수 있습니다 — AGENTS.md 규칙 4). `--area <폴더>`를 붙이면 그 폴더 안의 모든 저장소를 대상으로 같은 일을 하며, 원래 체크아웃이 사라진 피처도 읽을 수 있습니다. 저장소 config.md에 `index_style: obsidian`을 적으면 Obsidian 볼트용 frontmatter와 인라인 `#태그`를 덧붙입니다. 생성 시각은 절대 쓰지 않으므로 내용이 같으면 diff도 생기지 않습니다. 종료 코드는 `0` 찾음, `1` 없음, `2` 사용법 오류 또는 거부입니다.
 
diff --git a/README.md b/README.md
index 2d56793..5a41802 100644
--- a/README.md
+++ b/README.md
@@ -145,7 +145,7 @@ Per feature, inside the **target project**:
 │   ├── POLICY.md                     # human-declared hard rules; agents transcribe only
 │   ├── INDEX.md                      # ≤50 lines of lesson pointers
 │   ├── DOMAIN.md                     # terms · facts and constraints that span areas
-│   ├── areas/<area-slug>.md               # one per product area (web app: one menu): business rules P1… · numbers N1… (optional) · how it works · history
+│   ├── areas/<menu path>.md               # one per product area (web app: one menu), e.g. "학습 - 평가 - 제출.md": business rules P1… · numbers N1… (optional) · how it works · history · evidence (folded)
 │   └── lessons/<date>-<lesson>.md
 ├── work/<slug>/                      # OPEN features only
 │   ├── origin.md                     # the ticket / 기획서 as requested — bound by the intent gate
@@ -169,7 +169,7 @@ Per feature, inside the **target project**:
 
 **Where the records live is your choice.** By default they sit in the project's working copy. `init.sh . --area ~/knowledge` puts them in a folder you choose instead — `<area>/<unit>-<checkout-id>/`, with `.sdlc` linked to it, one store per checkout so two worktrees never share approvals. The area is refused if it sits inside the project (or the project inside it), if another checkout already owns that store, if a real `.sdlc` directory is already there (nothing is ever relocated for you), or if the link cannot be made. That ownership is re-checked at RUNTIME, not only at init: `check-gate.sh`, `approve.sh`, `close.sh`, `status.sh`, `tools/auto.sh`, `tools/verify.sh` and `tools/handoff.sh` refuse before any verdict or write when `<store>/PROJECT` names a different checkout, so a copied working copy (`cp -R`, rsync and most restores keep the symlink) can neither open another checkout's gate nor close its features. Reading is never bound that way: `tools/kb.sh show|search|list` still works, and nothing is ever re-bound or moved for you. Whichever you choose, **the store is yours to back up** — git no longer does it for you.
 
-**Knowledge is filed by product area.** For a web app an area is one menu, named by its menu path (`학습 > 평가 > 제출`); for other software a module, API, job, or CLI command. Each area has one page, `memory/areas/<area-slug>.md`: its business rules numbered P1, P2… in sentences a non-developer can read, how it works, and a history line per feature that changed it. An area that shows counts, rates, scores or charts also gets a Numbers section, N1, N2…, one line per figure recording what it counts and its source — the same permanence and treatment as a business rule. A feature's `summary.md` names its area on an `Area:` line, the spec states which rules or figures it keeps or changes, the Side effects verifier re-checks the ones it should not have touched, and the close merge files new or changed rules and figures on the page. `tools/kb.sh show "학습 > 평가 > 제출"` (or the page's file name) prints the page with the features that changed it.
+**Knowledge is filed by product area.** For a web app an area is one menu, named by its menu path (`학습 > 평가 > 제출`); for other software a module, API, job, or CLI command. Each area has one page, named by its menu path — `memory/areas/학습 - 평가 - 제출.md` (each ` > ` written ` - `; any of `/ \ : * ? " < > |` becomes `-`). The page reads reader first: its business rules numbered P1, P2… in sentences a non-developer can read, how it works, and a history line per feature that changed it; every source, verification, and code location sits in a folded evidence block at the bottom. An area that shows counts, rates, scores or charts also gets a Numbers section, N1, N2…, one line per figure recording what it counts and its source — the same permanence and treatment as a business rule. A feature's `summary.md` names its area on an `Area:` line, the spec states which rules or figures it keeps or changes, the Side effects verifier re-checks the ones it should not have touched, and the close merge files new or changed rules and figures on the page. `tools/kb.sh show "학습 > 평가 > 제출"` (or the page's file name) prints the page with the features that changed it.
 
 **Reading the records back** is `tools/kb.sh`: `index` regenerates the contents page (`init.sh` and `close.sh` do it for you) — the product areas with their rule count, last change and features, an overview table by state, date, area and tags, newest first, the harvests no close has merged yet, then one section per feature; `show <slug>` prints one feature as a digest — goal, its `summary.md` (area, what was wrong, before → after, how to check, the one record meant to be kept current), delivery, unmerged harvest candidates, lesson titles, then the paths; `search "<text>"` does a bounded literal search over open and closed features plus durable memory; `harvest [--stale <days>]` lists open features whose harvest.md is not in memory yet, with idle time (a stale one may be merged without closing — AGENTS.md rule 4); and `--area <folder>` does any of these across every store in that folder — including features whose checkout no longer exists. `index_style: obsidian` in the store's config.md adds frontmatter and inline `#tags` for a vault; no timestamp is ever written, so an unchanged page produces no diff. Exit codes: `0` found, `1` nothing found, `2` usage error or refusal.
 
diff --git a/gates/selftest.sh b/gates/selftest.sh
index aa602c0..583db19 100755
--- a/gates/selftest.sh
+++ b/gates/selftest.sh
@@ -67,6 +67,21 @@ LC_ALL=en_US.UTF-8 bash "$kit/tools/kb.sh" index >/dev/null
 grep -qF '| [명단 > 내보내기](memory/areas/roster.md) | 1 | — | f1 |' .sdlc/README.md || fail "area table wrong: $(grep '명단' .sdlc/README.md)"
 out=$(bash "$kit/tools/kb.sh" show "명단 > 내보내기")
 case "$out" in (*"P1: 현재 학기만"*) ;; (*) fail "show <menu path> did not print the rule: $out";; esac
+#    a page named by its menu path (" > " → " - "), reader first: plain rules on
+#    top, evidence in <details> at the bottom — evidence lines do not count as rules
+mkdir -p .sdlc/work/f3; printf -- '- Area: 교사 > 학생 > 학급 분석\n' > .sdlc/work/f3/summary.md
+printf -- '%s\n' '# Area: 교사 > 학생 > 학급 분석' '- Menu: 교사 > 학생 > 학급 분석' '## Business rules (정책)' \
+  '- P1: 자기 학급만 본다' '- P2: 전학생은 빠진다' '- ~~P3: 지난 학기도 보인다~~' '## History' '- 2026-09-01 f3 — 참여율 추가' \
+  '<details>' '<summary>근거 · 코드 위치 (개발자용)</summary>' '' '- Where: ClassAnalysis#get' \
+  '- P1 — source: 기획서 · set by f3' '- P3 — retired 2026-09-01 by f3: 정책 변경' '' '</details>' > ".sdlc/memory/areas/교사 - 학생 - 학급 분석.md"
+LC_ALL=en_US.UTF-8 bash "$kit/tools/kb.sh" index >/dev/null
+grep -qF '| [교사 > 학생 > 학급 분석](memory/areas/교사%20-%20학생%20-%20학급%20분석.md) | 2 | 2026-09-01 | f3 |' .sdlc/README.md \
+  || fail "menu-named page row wrong: $(grep '교사' .sdlc/README.md)"
+for q in "교사 > 학생 > 학급 분석" "교사 - 학생 - 학급 분석"; do
+  out=$(bash "$kit/tools/kb.sh" show "$q") || fail "show '$q' found no page"
+  case "$out" in (*"P1: 자기 학급만"*"근거 · 코드 위치 (개발자용):"*"Where: ClassAnalysis#get"*) ;; (*) fail "show '$q' not reader first: $out";; esac
+done
+bash "$kit/tools/kb.sh" show "../areas/roster" >/dev/null 2>&1 && fail "area lookup followed a path"
 echo "ok: knowledge by product area"
 
 echo "SELFTEST PASS"
diff --git a/init.sh b/init.sh
index ca585fb..b7cf26f 100755
--- a/init.sh
+++ b/init.sh
@@ -303,7 +303,7 @@ EOF
 
 [ -f .sdlc/memory/DOMAIN.md ] || cat > .sdlc/memory/DOMAIN.md <<'EOF'
 # Domain knowledge — how THIS system works: terms and facts that span areas
-# (≤100 lines; over → move area-specific facts to memory/areas/<area-slug>.md)
+# (≤100 lines; over → move area-specific facts to memory/areas/<menu path>.md)
 # ONE writer: the close step. Mid-loop candidates stage in the feature's
 # work/<slug>/harvest.md and merge here at close (AGENTS.md rule 4).
 # Facts carry [verified: how — YYYY-MM-DD]. Recency wins: a merge candidate
diff --git a/log/changelog-v0.16.0.md b/log/changelog-v0.16.0.md
index f0223f1..ab9f6a4 100644
--- a/log/changelog-v0.16.0.md
+++ b/log/changelog-v0.16.0.md
@@ -1,12 +1,16 @@
-# v0.16.0 — area pages record how a figure is calculated
+# v0.16.0 — area pages: figures, menu-path file names, reader-first layout
 
 v0.15.0 gave a product area one home for its business rules (P-numbers). A
 menu that shows counts, rates, scores or charts had nowhere to say what a
 number on screen actually means — "what is counted, out of what, real-time
 or batch" lived in someone's head or nowhere. This release adds an optional
 Numbers section to the area page, numbered N1… with the same permanence and
-filing rules as a business rule. Gates, approvals, the `--json` schema, and
-`kb.sh`'s rule count are unchanged.
+filing rules as a business rule. It also names each area page by its menu
+path, so the file list reads like the product's menus, and lays the page out
+reader first: plain language on top for a non-developer, every source,
+verification, and code location folded into an evidence block at the
+bottom. Gates, approvals, and the `--json` schema are unchanged; `kb.sh`'s
+rule count still counts only the live `- P<n>:` rules.
 
 ## Changes
 
@@ -31,12 +35,63 @@ filing rules as a business rule. Gates, approvals, the `--json` schema, and
   - AGENTS.md rule 4's "what goes where" table gains a row: a figure's
     calculation method files to the area page's Numbers, written by the
     close merge on a shipped close only — same as a business rule.
+- **An area page is named by its menu path.** The file name is the Menu
+  line with each ` > ` written ` - `, and any character a Windows or macOS
+  file name may not hold (`/ \ : * ? " < > |`) replaced with `-`: Menu
+  `교사 > 학생 > 학급 분석` → `memory/areas/교사 - 학생 - 학급 분석.md`. The
+  "ASCII kebab-case file name" instruction is gone from `templates/area.md`,
+  AGENTS.md rule 4, the READMEs, and `init.sh`'s DOMAIN.md header; the
+  close merge (AGENTS.md rule 4, `skills/5-ship`) is told the rule when it
+  creates a page.
+- **`tools/kb.sh`** finds an area page by its file name, by its Menu line,
+  or by the file name a Menu gives (`kb_area_file`); a name holding `/` or
+  `\`, or starting with `.`, is still never looked up as a file. The
+  contents page's Product areas links percent-encode space `%` `(` `)` `[`
+  `]` `#` `<` `>` (`kb_area_href`) so a link to `교사 - 학생 - 학급 분석.md`
+  opens in GitHub and Obsidian; Hangul bytes stay as written. `kb_body`
+  prints a `<summary>…</summary>` line as a section label, so `show` prints
+  the page top first and the evidence block last, under
+  `근거 · 코드 위치 (개발자용):` — the part the line bound cuts first. The
+  Rules column no longer counts an unfilled template line `- P1: <…>`,
+  and `show` drops such a line (`- <label with digits>: <…>`) the way it
+  already dropped `- Where: <…>`.
+- **Reader-first `templates/area.md`.** Menu and Aliases on top; Business
+  rules, Numbers, How it works, and History in plain sentences with no
+  source and no code; then a `<details>` block — "근거 · 코드 위치
+  (개발자용)" — holding the Where line and one evidence line per rule or
+  figure: `- P1 — source: … · set by <slug> · [verified: how — date]`. The
+  evidence line has no colon after the number, so it is never counted as a
+  rule. A retired rule stays struck through on top (`- ~~P3: …~~`) and its
+  evidence line carries `retired YYYY-MM-DD by <slug>: <why>`. The section
+  headings are unchanged. `templates/harvest.md`'s area candidates split
+  the plain sentence from its evidence the same way, and ship's
+  retrospective says which part goes where.
 - **README.md / README.ko.md**: the "Knowledge is filed by product area"
   paragraph and the memory-tree comment now mention the optional Numbers
-  section.
+  section, the menu-path file name, and the folded evidence block.
 
 ## Compatibility
 
+**Existing ASCII-named area pages keep working.** `kb.sh show "<menu
+path>"` still finds a page by its Menu line whatever the file is called,
+the contents page still lists and links it, and features still file under
+it by their `Area:` line. Only a newly created page takes the menu-path
+name. To rename an old page, move it to the name its Menu line gives (a
+plain `mv` when the store is not in git), then regenerate the contents page:
+
+```bash
+git mv ".sdlc/memory/areas/class-analysis.md" ".sdlc/memory/areas/교사 - 학생 - 학급 분석.md"
+tools/kb.sh index
+```
+ A feature whose `Area:` line names the old file name
+instead of the Menu should be changed to the Menu words. Nothing is renamed
+for you.
+
+**Existing pages with inline sources keep working.** A rule written the old
+way — `- P1: <rule> — source: … · set by …` — still counts and still
+prints; the close merge moves its source to the evidence block the next
+time it edits that rule, or you can move it by hand.
+
 An area page with no Numbers section is unaffected: the section is
 optional and `kb.sh`'s Rules column stays a P-number count only (a store
 with only P-numbers renders identically to before). No script parses the
@@ -45,7 +100,10 @@ with only P-numbers renders identically to before). No script parses the
 
 ## Validation
 
-- `bash gates/selftest.sh` → `SELFTEST PASS` (unchanged; the smoke test
-  does not exercise the Numbers section, and this change does not touch
-  the scripts it does cover — gate mechanics, lazymode, close proof,
-  product-area filing by menu path).
+- `bash gates/selftest.sh` → `SELFTEST PASS`. Its "knowledge by product
+  area" test now also files a feature under a Hangul menu page named
+  `교사 - 학생 - 학급 분석.md`, laid out reader first with evidence lines and
+  a retired rule: the contents-page row links the percent-encoded path with
+  a Rules count of 2 (evidence and retired lines not counted); `show` by
+  Menu and by file name prints the rules before the evidence block; a
+  lookup holding `/` is refused. With the old `kb.sh`, the test fails.
diff --git a/skills/5-ship/SKILL.md b/skills/5-ship/SKILL.md
index f036324..ca26125 100644
--- a/skills/5-ship/SKILL.md
+++ b/skills/5-ship/SKILL.md
@@ -66,12 +66,17 @@ no lesson. Write what there is into the feature's
 `.sdlc/work/<slug>/harvest.md` (lesson candidates in the skill 6 format;
 durable terms, verified facts, and constraints as domain candidates;
 AGENTS.md rule 4). **Every business rule this feature set, changed, or
-retired is an area candidate** — the rule in one testable sentence, its
-source (the origin, the spec R-item or compact intent O-item), the P-number
-it changes or retires — plus one
-history line per area it changed. **A figure the area page shows a Numbers
-section for is treated the same way**: the N-number it sets, changes, or
-retires, named on screen, with what it counts and its source. Then finish the feature's `summary.md`:
+retired is an area candidate** — the rule in one plain, testable sentence a
+non-developer can read, the P-number it changes or retires, and, as its
+evidence, the source (the origin, the spec R-item or compact intent O-item)
+and how it was verified — plus one history line per area it changed. **A
+figure the area page shows a Numbers section for is treated the same way**:
+the N-number it sets, changes, or retires, named on screen, with what it
+counts, and its source as evidence. The close merge writes the sentence on
+top of the area page and the evidence on the page's evidence line at the
+bottom; a page it creates is named by its menu path — the Menu with each
+` > ` written ` - ` and any of `/ \ : * ? " < > |` replaced with `-`
+(templates/area.md). Then finish the feature's `summary.md`:
 `Status` says only what delivery.md confirms (a pushed review branch is not
 a deployment), Before → After and How to check match what was actually
 built and proven, Remember holds the one thing worth knowing next time.
diff --git a/templates/area.md b/templates/area.md
index 2f7e3f3..b84d0d3 100644
--- a/templates/area.md
+++ b/templates/area.md
@@ -1,39 +1,59 @@
-# Area: <the area as users name it — for a web app the menu path, e.g. 학습 > 평가 > 제출>
+# Area: <menu path>
 
-<!-- One page per product area, at .sdlc/memory/areas/<area-slug>.md (ASCII
-     kebab-case file name; the human name lives on the Menu line). For a web
-     app an area is one menu — split a page when its sub-screens carry
-     different rules or it passes ~60 lines. Other software: a module, API,
-     job, or CLI command. This is the page a person or an agent opens to learn
-     what the area does and which business rules hold.
-     ONE writer: the close merge (AGENTS.md rule 4). Plain language a
-     non-developer can read; code identifiers only on the Where line. -->
+<!-- One page per product area, at .sdlc/memory/areas/<file name>.md. The file
+     name is the Menu line with each " > " written as " - ", and any of
+     / \ : * ? " < > | replaced with "-":
+       Menu 교사 > 학생 > 학급 분석 → memory/areas/교사 - 학생 - 학급 분석.md
+     For a web app an area is one menu — split a page when its sub-screens
+     carry different rules or it passes ~60 lines. Other software: a module,
+     API, job, or CLI command. ONE writer: the close merge (AGENTS.md rule 4).
+     READER FIRST: everything above <details> is for a non-developer — plain
+     sentences, no source, no code. Every source, date, verification, and code
+     identifier goes in the <details> block at the bottom, one evidence line
+     per rule or figure. -->
 
-- Menu: <the path as users see it; unique across pages, no commas — summary.md's Area line must read the same>
-- Where: <URL / route · entry point file:symbol · API endpoint · job name>
-- Aliases: <other words people search with — screen title, ticket wording, other language>
+- Menu: <menu path as users see it; no commas>
+- Aliases: <other words people search with>
+
+<!-- Menu is unique across pages and reads exactly as summary.md's Area line.
+     Aliases: screen title, ticket wording, other language. -->
 
 ## Business rules (정책)
-<!-- The product's rules for this area, one testable sentence each:
-       - P1: <rule> — source: <기획서 / ticket / human, date> · set by <slug> · [verified: how — YYYY-MM-DD]
+<!-- The product's rules for this area, one plain sentence each, testable,
+     nothing else on the line. Its source goes on its evidence line below.
      P-numbers are permanent. The close merge gives a new rule the next number
      this page has never used (retired ones count), edits a changed rule in
-     place with its new source, and keeps a retired one struck through:
-       - ~~P3: <rule>~~ — retired YYYY-MM-DD by <slug>: <why>
+     place (and its evidence line), and keeps a retired one struck through:
+       - ~~P3: <rule>~~
+     with the reason on its evidence line:
+       - P3 — retired YYYY-MM-DD by <slug>: <why> · source: <…>
      Not POLICY.md — that file holds the human's rules for agents. -->
+- P1: <rule in one plain sentence — no source, no code>
 
 ## Numbers (통계 산정)
 <!-- Only for an area that shows counts, rates, scores or charts; delete the
-     section otherwise. One line per figure, named as the screen labels it:
-       - N1: <label on screen> — <what is counted> / <out of what> · <real-time | batch: <job>, refreshed <when>, data up to <when>> — source: <기획서 / ticket / code, date> · set by <slug> · [verified: how — YYYY-MM-DD]
+     section otherwise. One line per figure, named as the screen labels it.
      N-numbers follow the P-number rules above: permanent, a changed figure is
-     edited in place, a retired one is kept struck through. -->
+     edited in place, a retired one is kept struck through with the reason on
+     its evidence line. The batch job's name goes on the evidence line. -->
+- N1: <label on screen> — <what is counted> / <out of what> · <real-time | batch: refreshed <when>, data up to <when>>
 
 ## How it works
-<!-- ≤10 lines: what the user does, step by step, what the system does in
-     response, and facts that hold for this area only. Use DOMAIN.md terms;
-     do not redefine them here. -->
+<!-- What the user does, step by step, what the system does in response, and
+     facts that hold for this area only. Use DOMAIN.md terms; do not redefine
+     them here. -->
+<≤10 plain lines>
 
 ## History
-<!-- Newest first; one line per shipped feature that changed this area:
-       - YYYY-MM-DD <slug> — <what changed for the user, from its summary.md> -->
+<!-- Newest first; one line per shipped feature that changed this area, from
+     its summary.md. -->
+- YYYY-MM-DD <slug> — <what changed for the user>
+
+<details>
+<summary>근거 · 코드 위치 (개발자용)</summary>
+
+- Where: <route · file:symbol · endpoint · job>
+- P1 — source: <기획서 / ticket / human / code, date> · set by <slug> · [verified: how — YYYY-MM-DD]
+- N1 — source: … · set by … · [verified: …]
+
+</details>
diff --git a/templates/harvest.md b/templates/harvest.md
index 4c43802..b8c7437 100644
--- a/templates/harvest.md
+++ b/templates/harvest.md
@@ -6,13 +6,15 @@
      file exists. -->
 
 ## Area candidates
-<!-- - [area: <menu path>] rule new | P<n> changed | P<n> retired: <rule> — source · [verified: how — YYYY-MM-DD]
-     - [area: <menu path>] figure new | N<n> changed | N<n> retired: <label on screen> — what it counts · source · [verified: how — YYYY-MM-DD]
+<!-- - [area: <menu path>] rule new | P<n> changed | P<n> retired: <rule, one plain sentence> — evidence: source · [verified: how — YYYY-MM-DD] · why, if retired
+     - [area: <menu path>] figure new | N<n> changed | N<n> retired: <label on screen> — what it counts — evidence: source · [verified: how — YYYY-MM-DD]
      - [area: <menu path>] fact: <holds for this area only> — [verified: how — YYYY-MM-DD]
      - [area: <menu path>] history: YYYY-MM-DD <slug> — <what changed for the user>
      Rules, figures, and history merge only when the feature closes shipped;
      a stale merge leaves them here (rule 4). The merge numbers a new rule
-     or figure. -->
+     or figure, writes the plain sentence on top of the area page and the
+     evidence part on its evidence line at the bottom (templates/area.md).
+     Code identifiers belong only in the evidence part. -->
 
 ## Domain candidates
 <!-- - fact / term / constraint that spans areas — [verified: how — YYYY-MM-DD]
diff --git a/tools/kb.sh b/tools/kb.sh
index 0a6e67a..767ee2d 100755
--- a/tools/kb.sh
+++ b/tools/kb.sh
@@ -24,9 +24,11 @@
 #
 # `show` is a digest — goal, summary.md (templates/summary.md), delivery,
 # unmerged harvest.md candidates, lesson titles, then the paths. A name that is
-# no feature is looked up as a product area: memory/areas/<name>.md, or the page
-# whose Menu line reads exactly <name> (templates/area.md), printed with the
-# features whose summary.md names it. `index` opens with the product areas
+# no feature is looked up as a product area: memory/areas/<name>.md, else the
+# page whose Menu line reads exactly <name>, else the page whose file name that
+# Menu gives (" > " → " - ", templates/area.md), printed top first — the plain
+# rules before the developer evidence — with the features whose summary.md
+# names it. `index` opens with the product areas
 # (business-rule count, last change, features), then an overview table (state,
 # date, area, tags) newest first, and lists the harvests no close has merged yet. `index_style: obsidian` in the store's config.md adds
 # YAML frontmatter and inline #tags to the page; links stay relative markdown.
@@ -98,6 +100,8 @@ kb_field() { # <file> <label> → the first "- <label>: value" line's value (raw
   [ -f "$1" ] || return 1
   awk -v k="- $2:" 'index($0, k) == 1 { sub(/^[^:]*: */, ""); sub(/[ \t\r]*$/, ""); print; exit }' "$1"
 }
+# A field is read wherever its "- <label>:" line sits in the file — an area
+# page's Where line lives inside its <details> evidence block (templates/area.md).
 kb_get() { # <file> <label> → the value, or empty when absent or still a template placeholder (<…>, YYYY-…)
   local v; v=$(kb_field "$1" "$2" 2>/dev/null || true)
   case "$v" in ''|'<'*|YYYY*) printf '';; *) printf '%s' "$v";; esac
@@ -138,7 +142,8 @@ kb_hashtags() { # "tag, tag" → "#tag #tag" for Obsidian's tag pane (spaces ins
 }
 # The readable body of a record: template comments (single- and multi-line),
 # blank lines and unfilled `<placeholder>` lines are dropped; the H1 is dropped;
-# `## X` (and deeper headings) become `X:` labels, so a record embedded in the
+# `## X` (and deeper headings) and a `<summary>X</summary>` line (the area
+# page's evidence block) become `X:` labels, so a record embedded in the
 # contents page cannot hijack its outline. A heading with nothing under it yet
 # is dropped too: an unfinished record prints only what is known.
 kb_body() { # <file>
@@ -156,10 +161,11 @@ kb_body() { # <file>
       sub(/[ \t\r]+$/, "", line)
       if (line ~ /^# /) next
       if (line ~ /^##+ /) { sub(/^#+ +/, "", line); pend = line ":"; next }
+      if (line ~ /^[ \t]*<summary>.*<\/summary>[ \t]*$/) { sub(/^[ \t]*<summary>[ \t]*/, "", line); sub(/[ \t]*<\/summary>[ \t]*$/, "", line); pend = line ":"; next }
       if (line ~ /^[ \t]*$/) next
       if (line ~ /^[ \t]*[-*] *<[^>]*>[ \t]*$/) next
       if (line ~ /^[ \t]*<[^>]*>[ \t]*$/) next
-      if (line ~ /^[ \t]*[-*] *[A-Za-z-]+: *<[^>]*>[ \t]*$/) next
+      if (line ~ /^[ \t]*[-*] *[A-Za-z0-9-]+: *<[^>]*>[ \t]*$/) next
       if (pend != "") { print pend; pend = "" }
       print line
     }' "$1"
@@ -215,7 +221,10 @@ kb_area_menu() { # <page> → its Menu line, else its H1, else its file name (an
   case "$m" in ''|'<'*) m=$(basename "$1" .md);; esac
   printf '%s' "$m"
 }
-kb_area_rules() { awk '/^- P[0-9]+:/ { n++ } END { print n + 0 }' "$1"; }   # live rules; a retired one reads "- ~~P…"
+# Live rules: the top "- P<n>: rule" lines. A retired one reads "- ~~P…", an
+# evidence line "- P<n> — source…" (no colon after the number), and an unfilled
+# template line "- P1: <…>" — none of them counts.
+kb_area_rules() { awk '/^- P[0-9]+:/ && !/^- P[0-9]+: *</ { n++ } END { print n + 0 }' "$1"; }
 kb_area_last() { # <page> → the date of the newest History line
   awk '/^## /{ h = ($0 ~ /^## History/); next } h && /^- [0-9][0-9][0-9][0-9]-[0-9][0-9]-[0-9][0-9]/ { print substr($0, 3, 10); exit }' "$1"
 }
@@ -230,15 +239,33 @@ kb_area_of() { # <menu> <file name>; kb_area_refs lines on stdin → the lines n
   awk -F "$SEP" -v a="$1" -v b="$2" '$1 == a || $1 == b'
 }
 kb_area_slugs() { awk -F "$SEP" '{ if (n++) printf ", "; printf "%s", $3 }'; }   # kb_area_refs lines → "slug, slug"
-kb_area_find() { # <store> <name> → the page: <name>.md, else the page whose Menu is <name>
-  local f
-  case "$2" in */*|.*) ;; *)
-    f="$1/memory/areas/$2.md"
-    [ -f "$f" ] && [ ! -L "$f" ] && { printf '%s\n' "$f"; return 0; };;
-  esac
-  kb_area_pages "$1" | while IFS= read -r f; do
+# The file name a Menu gives its page (templates/area.md): each " > " written
+# " - ", and every character a Windows or macOS file name may not hold
+# (/ \ : * ? " < > |) replaced with "-". Byte-wise under LC_ALL=C: no byte of a
+# multibyte (Hangul) character falls in an ASCII class, so it passes unchanged.
+kb_area_file() { printf '%s' "$1" | sed -e 's/ > / - /g' -e 's#[/\\:*?"<>|]#-#g'; }
+kb_area_page() { # <store> <file name, no .md> → the page when it exists; a name with / or \ or a leading . is refused
+  case "$2" in ''|*/*|*\\*|.*) return 1;; esac
+  [ -f "$1/memory/areas/$2.md" ] && [ ! -L "$1/memory/areas/$2.md" ] && printf '%s\n' "$1/memory/areas/$2.md"
+}
+kb_area_find() { # <store> <name> → the page: <name>.md, else the page whose Menu is <name>, else the page named by that Menu
+  local f d
+  kb_area_page "$1" "$2" && return 0
+  f=$(kb_area_pages "$1" | while IFS= read -r f; do
     [ "$(kb_area_menu "$f")" = "$2" ] && { printf '%s\n' "$f"; break; }
-  done
+  done)
+  [ -n "$f" ] && { printf '%s\n' "$f"; return 0; }
+  d=$(kb_area_file "$2")
+  [ "$d" = "$2" ] || kb_area_page "$1" "$d"
+  return 0
+}
+# A page's link target on the contents page. File names hold spaces (and, from a
+# Menu, Hangul): space % ( ) [ ] # < > are percent-encoded so the link works in
+# GitHub and Obsidian alike; other UTF-8 bytes stay as they are, the form
+# Obsidian writes its own markdown links in.
+kb_area_href() { # <file name> → memory/areas/<encoded>
+  printf 'memory/areas/%s' "$(printf '%s' "$1" | sed -e 's/%/%25/g' -e 's/ /%20/g' -e 's/(/%28/g' -e 's/)/%29/g' \
+    -e 's/\[/%5B/g' -e 's/\]/%5D/g' -e 's/#/%23/g' -e 's/</%3C/g' -e 's/>/%3E/g')"
 }
 # "<slug>\t<lines>\t<dir>" for every open feature whose harvest.md still holds
 # candidates — the one definition `index` and `harvest` share.
@@ -321,7 +348,7 @@ kb_index() { # <store>
         menu=$(kb_area_menu "$f")
         body=$(printf '%s\n' "$refs" | kb_area_of "$menu" "$(basename "$f" .md)" | kb_area_slugs)
         f_last=$(kb_area_last "$f")
-        printf '| [%s](memory/areas/%s) | %s | %s | %s |\n' "$(kb_table_cell "$menu")" "$(basename "$f")" \
+        printf '| [%s](%s) | %s | %s | %s |\n' "$(kb_table_cell "$menu")" "$(kb_area_href "$(basename "$f")")" \
           "$(kb_area_rules "$f")" "${f_last:-—}" "${body:-—}"
       done
       printf '%s\n' "$refs" | awk -F "$SEP" 'NF && !seen[$1]++ { print $1 }' | while IFS= read -r menu; do

From a363d7a2f1e397fd19ec20705ad013ed85fb1dba Mon Sep 17 00:00:00 2001
From: "sungkuk.chae-at-471112812288"
 <76669236+cskwork@users.noreply.github.com>
Date: Wed, 23 Sep 2026 13:00:26 +0900
Subject: [PATCH 3/4] feat: area evidence block also as an Obsidian folded
 callout (v0.16.0)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Obsidian does not render markdown inside HTML blocks, so a <details>
evidence block collapses into one raw paragraph there. A store with
index_style: obsidian writes it as a folded callout instead:
  > [!info]- 근거 · 코드 위치 (개발자용)

- templates/area.md: <details> stays the default; one comment names the
  callout form for an Obsidian store
- tools/kb.sh: kb_field/kb_get drop a leading "> " so Where is found in a
  callout; kb_body treats a "> [!type]- X" line as the "X:" label and
  prints the callout's lines without "> "; kb_area_rules still counts only
  column-0 "- P<n>:" lines
- AGENTS.md, skills/5-ship, templates/harvest.md: "or the folded callout in
  an Obsidian store (index_style: obsidian)"
- selftest: a callout-form page (Where read by kb_get, rule count 1 with a
  "> - P2:" line inside, evidence printed last without the prefix)
- changelog v0.16.0: one entry

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
---
 AGENTS.md                |  5 +++--
 gates/selftest.sh        | 12 ++++++++++++
 log/changelog-v0.16.0.md |  8 +++++++-
 skills/5-ship/SKILL.md   |  5 +++--
 templates/area.md        |  5 +++++
 templates/harvest.md     |  4 +++-
 tools/kb.sh              | 17 +++++++++++------
 7 files changed, 44 insertions(+), 12 deletions(-)

diff --git a/AGENTS.md b/AGENTS.md
index ae98475..6f67b5e 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -149,7 +149,7 @@ archives the feature to `.sdlc/archive/<slug>/`.
    |---|---|---|
    | A product's business rule (정책) — "a submitted answer cannot be edited" | `memory/areas/<menu path>.md` Business rules, numbered P1…, one plain sentence each (templates/area.md) | close merge, shipped only |
    | How an on-screen count, rate, score or chart figure is calculated | the same page's Numbers, numbered N1… (templates/area.md) | close merge, shipped only |
-   | Where a rule or figure came from, and the code behind the area | the same page's evidence block at the bottom (`<details>`): `P1 — source · set by · [verified]`, the Where line | close merge |
+   | Where a rule or figure came from, and the code behind the area | the same page's evidence block at the bottom (`<details>`, or the folded callout `> [!info]-` in an Obsidian store (index_style: obsidian)): `P1 — source · set by · [verified]`, the Where line | close merge |
    | What changed in an area, when | the area page's History, one line per feature | close merge, shipped only |
    | A fact that holds for one area only | the area page's How it works | close merge |
    | A term, or a system fact/constraint that spans areas | `memory/DOMAIN.md` (100 lines max) | close merge |
@@ -166,7 +166,8 @@ archives the feature to `.sdlc/archive/<slug>/`.
    (`교사 > 학생 > 학급 분석` → `memory/areas/교사 - 학생 - 학급 분석.md`).
    An area page reads reader first: rules, figures, how it works and history
    in plain language on top; every source, verification, and code identifier
-   in the evidence block at the bottom. DOMAIN over its limit: move
+   in the evidence block at the bottom — a `<details>` block, or the folded
+   callout in an Obsidian store (index_style: obsidian). DOMAIN over its limit: move
    area-specific facts to their area page. INDEX over its limit: merge
    near-duplicates, drop superseded entries, replace promoted ones
    (skills/6-maintain).
diff --git a/gates/selftest.sh b/gates/selftest.sh
index 583db19..d15e57e 100755
--- a/gates/selftest.sh
+++ b/gates/selftest.sh
@@ -82,6 +82,18 @@ for q in "교사 > 학생 > 학급 분석" "교사 - 학생 - 학급 분석"; do
   case "$out" in (*"P1: 자기 학급만"*"근거 · 코드 위치 (개발자용):"*"Where: ClassAnalysis#get"*) ;; (*) fail "show '$q' not reader first: $out";; esac
 done
 bash "$kit/tools/kb.sh" show "../areas/roster" >/dev/null 2>&1 && fail "area lookup followed a path"
+#    the Obsidian form of the evidence block: a folded callout, lines prefixed "> "
+printf -- '%s\n' '- Menu: 학생 > 과제' '## Business rules (정책)' '- P1: 마감 후 제출 불가' '## History' '- 2026-09-02 f4 — 마감 표시' \
+  '> [!info]- 근거 · 코드 위치 (개발자용)' '> - Where: HomeworkApi#submit' '> - P1 — source: 기획서 · set by f4' '> - P2: 콜론 줄도 규칙이 아니다' \
+  > ".sdlc/memory/areas/학생 - 과제.md"
+LC_ALL=en_US.UTF-8 bash "$kit/tools/kb.sh" index >/dev/null
+grep -qF '| [학생 > 과제](memory/areas/학생%20-%20과제.md) | 1 | 2026-09-02 | — |' .sdlc/README.md || fail "callout page row wrong: $(grep '과제' .sdlc/README.md)"
+out=$(bash "$kit/tools/kb.sh" show "학생 > 과제") || fail "show found no callout page"
+case "$out" in (*"P1: 마감 후"*"History:"*"근거 · 코드 위치 (개발자용):"*"  - Where: HomeworkApi#submit"*) ;; (*) fail "callout evidence not read last: $out";; esac
+case "$out" in (*"> -"*) fail "callout prefix printed: $out";; esac
+fns=$(sed -n -e '/^kb_field() {/,/^}/p' -e '/^kb_get() {/,/^}/p' "$kit/tools/kb.sh")
+where=$(eval "$fns"; kb_get ".sdlc/memory/areas/학생 - 과제.md" Where)
+[ "$where" = "HomeworkApi#submit" ] || fail "kb_get did not read Where inside the callout: '$where'"
 echo "ok: knowledge by product area"
 
 echo "SELFTEST PASS"
diff --git a/log/changelog-v0.16.0.md b/log/changelog-v0.16.0.md
index ab9f6a4..d8eebaa 100644
--- a/log/changelog-v0.16.0.md
+++ b/log/changelog-v0.16.0.md
@@ -66,6 +66,11 @@ rule count still counts only the live `- P<n>:` rules.
   headings are unchanged. `templates/harvest.md`'s area candidates split
   the plain sentence from its evidence the same way, and ship's
   retrospective says which part goes where.
+- **Obsidian form of the evidence block.** Obsidian does not render markdown
+  inside `<details>`, so a store with `index_style: obsidian` writes it as
+  a folded callout (`> [!info]- 근거 · 코드 위치 (개발자용)`, every line
+  prefixed `> `); `kb.sh` reads both forms (Where found, callout lines
+  printed last without `> `, never counted as rules).
 - **README.md / README.ko.md**: the "Knowledge is filed by product area"
   paragraph and the memory-tree comment now mention the optional Numbers
   section, the menu-path file name, and the folded evidence block.
@@ -106,4 +111,5 @@ with only P-numbers renders identically to before). No script parses the
   a retired rule: the contents-page row links the percent-encoded path with
   a Rules count of 2 (evidence and retired lines not counted); `show` by
   Menu and by file name prints the rules before the evidence block; a
-  lookup holding `/` is refused. With the old `kb.sh`, the test fails.
+  lookup holding `/` is refused; a second page in the Obsidian callout form
+  is read the same way. With the old `kb.sh`, the test fails.
diff --git a/skills/5-ship/SKILL.md b/skills/5-ship/SKILL.md
index ca26125..0f3899d 100644
--- a/skills/5-ship/SKILL.md
+++ b/skills/5-ship/SKILL.md
@@ -73,8 +73,9 @@ and how it was verified — plus one history line per area it changed. **A
 figure the area page shows a Numbers section for is treated the same way**:
 the N-number it sets, changes, or retires, named on screen, with what it
 counts, and its source as evidence. The close merge writes the sentence on
-top of the area page and the evidence on the page's evidence line at the
-bottom; a page it creates is named by its menu path — the Menu with each
+top of the area page and the evidence on the page's evidence line in the
+`<details>` block at the bottom (or the folded callout in an Obsidian store,
+index_style: obsidian); a page it creates is named by its menu path — the Menu with each
 ` > ` written ` - ` and any of `/ \ : * ? " < > |` replaced with `-`
 (templates/area.md). Then finish the feature's `summary.md`:
 `Status` says only what delivery.md confirms (a pushed review branch is not
diff --git a/templates/area.md b/templates/area.md
index b84d0d3..7ebce2c 100644
--- a/templates/area.md
+++ b/templates/area.md
@@ -49,6 +49,11 @@
      its summary.md. -->
 - YYYY-MM-DD <slug> — <what changed for the user>
 
+<!-- In a store with `index_style: obsidian` write this block as a folded
+     callout instead, every line prefixed "> ":
+       > [!info]- 근거 · 코드 위치 (개발자용)
+       > - Where: <…>
+       > - P1 — source: … · set by … · [verified: …] -->
 <details>
 <summary>근거 · 코드 위치 (개발자용)</summary>
 
diff --git a/templates/harvest.md b/templates/harvest.md
index b8c7437..5966d40 100644
--- a/templates/harvest.md
+++ b/templates/harvest.md
@@ -13,7 +13,9 @@
      Rules, figures, and history merge only when the feature closes shipped;
      a stale merge leaves them here (rule 4). The merge numbers a new rule
      or figure, writes the plain sentence on top of the area page and the
-     evidence part on its evidence line at the bottom (templates/area.md).
+     evidence part on its evidence line in the <details> block at the bottom,
+     or the folded callout in an Obsidian store (index_style: obsidian;
+     templates/area.md).
      Code identifiers belong only in the evidence part. -->
 
 ## Domain candidates
diff --git a/tools/kb.sh b/tools/kb.sh
index 767ee2d..1968c28 100755
--- a/tools/kb.sh
+++ b/tools/kb.sh
@@ -96,9 +96,9 @@ kb_stores() { # <store-or-empty> <area-or-empty>
 # --- record fields -----------------------------------------------------------
 KB_DOCS="summary.md origin.md intent.md spec.md plan.md map.md evidence.md delivery.md CLOSED"
 
-kb_field() { # <file> <label> → the first "- <label>: value" line's value (raw)
+kb_field() { # <file> <label> → the first "- <label>: value" line's value (raw); a "> " quote prefix (an Obsidian callout) is dropped first
   [ -f "$1" ] || return 1
-  awk -v k="- $2:" 'index($0, k) == 1 { sub(/^[^:]*: */, ""); sub(/[ \t\r]*$/, ""); print; exit }' "$1"
+  awk -v k="- $2:" '{ sub(/^> ?/, "") } index($0, k) == 1 { sub(/^[^:]*: */, ""); sub(/[ \t\r]*$/, ""); print; exit }' "$1"
 }
 # A field is read wherever its "- <label>:" line sits in the file — an area
 # page's Where line lives inside its <details> evidence block (templates/area.md).
@@ -142,14 +142,16 @@ kb_hashtags() { # "tag, tag" → "#tag #tag" for Obsidian's tag pane (spaces ins
 }
 # The readable body of a record: template comments (single- and multi-line),
 # blank lines and unfilled `<placeholder>` lines are dropped; the H1 is dropped;
-# `## X` (and deeper headings) and a `<summary>X</summary>` line (the area
-# page's evidence block) become `X:` labels, so a record embedded in the
+# `## X` (and deeper headings), a `<summary>X</summary>` line and a
+# `> [!type]- X` callout line (the area page's evidence block, HTML or
+# Obsidian form) become `X:` labels — the callout's own lines print without
+# their "> " prefix — so a record embedded in the
 # contents page cannot hijack its outline. A heading with nothing under it yet
 # is dropped too: an unfinished record prints only what is known.
 kb_body() { # <file>
   [ -f "$1" ] || return 0
   awk '
-    BEGIN { c = 0 }
+    BEGIN { c = 0; q = 0 }
     {
       line = $0
       if (c) { i = index(line, "-->"); if (i) { c = 0; line = substr(line, i + 3) } else next }
@@ -159,6 +161,8 @@ kb_body() { # <file>
         else { line = substr(line, 1, s - 1); c = 1; break }
       }
       sub(/[ \t\r]+$/, "", line)
+      if (q) { if (line ~ /^>/) sub(/^> ?/, "", line); else q = 0 }
+      if (line ~ /^> *\[![A-Za-z-]+\][-+]?/) { sub(/^> *\[![A-Za-z-]+\][-+]? */, "", line); pend = line ":"; q = 1; next }
       if (line ~ /^# /) next
       if (line ~ /^##+ /) { sub(/^#+ +/, "", line); pend = line ":"; next }
       if (line ~ /^[ \t]*<summary>.*<\/summary>[ \t]*$/) { sub(/^[ \t]*<summary>[ \t]*/, "", line); sub(/[ \t]*<\/summary>[ \t]*$/, "", line); pend = line ":"; next }
@@ -222,7 +226,8 @@ kb_area_menu() { # <page> → its Menu line, else its H1, else its file name (an
   printf '%s' "$m"
 }
 # Live rules: the top "- P<n>: rule" lines. A retired one reads "- ~~P…", an
-# evidence line "- P<n> — source…" (no colon after the number), and an unfilled
+# evidence line "- P<n> — source…" (no colon after the number; "> - P<n> —" in
+# an Obsidian callout — never at column 0), and an unfilled
 # template line "- P1: <…>" — none of them counts.
 kb_area_rules() { awk '/^- P[0-9]+:/ && !/^- P[0-9]+: *</ { n++ } END { print n + 0 }' "$1"; }
 kb_area_last() { # <page> → the date of the newest History line

From cde44ff0f87d7d33612fa6d0b5d5f7802b37974b Mon Sep 17 00:00:00 2001
From: "sungkuk.chae-at-471112812288"
 <76669236+cskwork@users.noreply.github.com>
Date: Wed, 23 Sep 2026 13:11:27 +0900
Subject: [PATCH 4/4] feat: area pages use tables for rules, figures, history
 and evidence (v0.16.0)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

- templates/area.md: Business rules | # | Rule |, Numbers | # | Label on
  screen | What is counted | Out of | Timing |, History | Date | Feature |
  What changed for the user |, and in the evidence block (<details> or the
  Obsidian callout) the - Where: line plus | # | Source | Set by | Verified |.
  Menu, Aliases and How it works stay list lines; a retired rule is
  | ~~P3~~ | ~~rule~~ | with the reason in its evidence row's Source cell;
  header words are free, a literal | in a cell is \|
- tools/kb.sh keys on row shapes only: kb_area_rules counts | P<n> | rows
  and old - P<n>: lines above the evidence block (stops at <details> or
  > [!...]); kb_area_last reads History's first | YYYY-MM-DD | row or the
  old list line; kb_body prints tables as written (callout rows without
  "> "), dropping only template placeholder rows and the header of a table
  left with no rows
- AGENTS.md rule 4, skills/5-ship, templates/harvest.md: the close merge
  writes rows (rule/figure/history and its evidence row); harvest
  candidates stay line-shaped
- README.md / README.ko.md: one mention each
- selftest: table-form pages in <details> and callout form with Korean
  headers (evidence, retired and placeholder rows not counted, last change
  from the History table, show prints tables with evidence last); the
  list-form page still counts
- changelog v0.16.0: tables entry + list-form compatibility note

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
---
 AGENTS.md                | 19 ++++++------
 README.ko.md             |  2 +-
 README.md                |  2 +-
 gates/selftest.sh        | 37 ++++++++++++----------
 log/changelog-v0.16.0.md | 53 ++++++++++++++++++++++---------
 skills/5-ship/SKILL.md   |  6 ++--
 templates/area.md        | 67 ++++++++++++++++++++++++----------------
 templates/harvest.md     |  5 +--
 tools/kb.sh              | 45 +++++++++++++++++++++------
 9 files changed, 154 insertions(+), 82 deletions(-)

diff --git a/AGENTS.md b/AGENTS.md
index 6f67b5e..1eade42 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -147,10 +147,10 @@ archives the feature to `.sdlc/archive/<slug>/`.
 
    | What | Where | Written by |
    |---|---|---|
-   | A product's business rule (정책) — "a submitted answer cannot be edited" | `memory/areas/<menu path>.md` Business rules, numbered P1…, one plain sentence each (templates/area.md) | close merge, shipped only |
-   | How an on-screen count, rate, score or chart figure is calculated | the same page's Numbers, numbered N1… (templates/area.md) | close merge, shipped only |
-   | Where a rule or figure came from, and the code behind the area | the same page's evidence block at the bottom (`<details>`, or the folded callout `> [!info]-` in an Obsidian store (index_style: obsidian)): `P1 — source · set by · [verified]`, the Where line | close merge |
-   | What changed in an area, when | the area page's History, one line per feature | close merge, shipped only |
+   | A product's business rule (정책) — "a submitted answer cannot be edited" | `memory/areas/<menu path>.md` Business rules table, one row per rule — P1…, one plain sentence (templates/area.md) | close merge, shipped only |
+   | How an on-screen count, rate, score or chart figure is calculated | the same page's Numbers table, one row per figure — N1… (templates/area.md) | close merge, shipped only |
+   | Where a rule or figure came from, and the code behind the area | the same page's evidence block at the bottom (`<details>`, or the folded callout `> [!info]-` in an Obsidian store (index_style: obsidian)): the Where line, then one evidence row per rule or figure — `\| P1 \| source \| set by \| verified \|` | close merge |
+   | What changed in an area, when | the area page's History table, one row per feature | close merge, shipped only |
    | A fact that holds for one area only | the area page's How it works | close merge |
    | A term, or a system fact/constraint that spans areas | `memory/DOMAIN.md` (100 lines max) | close merge |
    | A trap and the correct move | `memory/lessons/<date>-<slug>.md` + one `INDEX.md` line (50 lines max) | close merge |
@@ -164,8 +164,8 @@ archives the feature to `.sdlc/archive/<slug>/`.
    which is how `kb.sh` links them. The page's file name is that Menu with
    each ` > ` written ` - ` and any of `/ \ : * ? " < > |` replaced with `-`
    (`교사 > 학생 > 학급 분석` → `memory/areas/교사 - 학생 - 학급 분석.md`).
-   An area page reads reader first: rules, figures, how it works and history
-   in plain language on top; every source, verification, and code identifier
+   An area page reads reader first: rules, figures and history as table rows
+   and how it works as bullets, in plain language, on top; every source, verification, and code identifier
    in the evidence block at the bottom — a `<details>` block, or the folded
    callout in an Obsidian store (index_style: obsidian). DOMAIN over its limit: move
    area-specific facts to their area page. INDEX over its limit: merge
@@ -176,9 +176,10 @@ archives the feature to `.sdlc/archive/<slug>/`.
    `.sdlc/work/<slug>/harvest.md` (templates/harvest.md). At close, merge
    every candidate into its row above — creating an area page from the
    template when none exists (file named by its menu path, as above),
-   numbering new rules, putting each rule's or figure's source on its
-   evidence line, adding the feature's History line to each area it
-   changed — then delete harvest.md. Business
+   numbering new rules, writing each rule or figure as a table row and its
+   source as its evidence row, adding the feature's History row to each
+   area it changed (a line-shaped candidate becomes a row) — then delete
+   harvest.md. Business
    rules and History describe what the product DOES, so they merge only when
    the feature closes `shipped`: any other close drops them (summary.md keeps
    the story), and a stale merge leaves them in harvest.md for that close;
diff --git a/README.ko.md b/README.ko.md
index 3bad008..7ac45d1 100644
--- a/README.ko.md
+++ b/README.ko.md
@@ -169,7 +169,7 @@ agent  APPROVED: intent of claims-status (.sdlc/work/claims-status/intent.md)
 
 **기록을 어디에 둘지는 사용자가 정합니다.** 기본값은 프로젝트 작업 사본 안이고, `init.sh . --area ~/knowledge`를 쓰면 사용자가 고른 폴더 아래 `<area>/<단위이름>-<체크아웃 식별자>/`에 저장하고 `.sdlc`를 그곳으로 연결합니다. 체크아웃마다 저장소가 하나씩이므로 워크트리 두 개가 승인 상태를 공유하는 일이 없습니다. 영역이 프로젝트 안에 있거나 프로젝트가 영역 안에 있을 때, 다른 체크아웃이 이미 그 저장소를 소유할 때, 실제 `.sdlc` 디렉터리가 이미 있을 때(자동으로 옮기지 않습니다), 링크를 만들 수 없을 때는 아무것도 쓰지 않고 분명히 실패합니다. 이 소유권은 init 시점뿐 아니라 실행 시점에도 다시 확인합니다. `<store>/PROJECT`에 적힌 체크아웃이 지금 실행 중인 체크아웃과 다르면 `check-gate.sh`, `approve.sh`, `close.sh`, `status.sh`, `tools/auto.sh`, `tools/verify.sh`, `tools/handoff.sh`가 판정을 내리거나 상태를 쓰기 전에 거부하므로, 심볼릭 링크를 그대로 복사한 작업 사본(`cp -R`, rsync, 대부분의 백업 복원)이 다른 체크아웃의 게이트를 열거나 그 피처를 닫을 수 없습니다. 읽기는 이 제약을 받지 않아 `tools/kb.sh show|search|list`는 그대로 쓸 수 있고, 소유권을 자동으로 옮기거나 다시 묶는 일은 없습니다. 어느 쪽을 고르든 **저장소 백업은 사용자의 몫입니다.** git이 더 이상 대신해 주지 않습니다.
 
-**지식은 제품 영역별로 정리됩니다.** 웹앱이면 제품 영역은 메뉴 하나이고 메뉴 경로(`학습 > 평가 > 제출`)로 부릅니다. 다른 소프트웨어는 모듈, API, 배치 작업, CLI 명령입니다. 제품 영역마다 페이지가 한 장 있고, 파일 이름은 메뉴 경로입니다 — `memory/areas/학습 - 평가 - 제출.md`(` > `는 ` - `로, `/ \ : * ? " < > |`는 `-`로 바꿉니다). 페이지는 읽는 사람 우선입니다. 위에는 개발자가 아니어도 읽을 수 있는 문장으로 쓴 업무 정책(P1, P2…), 동작 방식, 그 영역을 바꾼 피처별 이력 한 줄이 오고, 출처·검증·코드 위치는 모두 맨 아래 접힌 근거 블록에 둡니다. 건수·비율·점수·차트를 보여주는 영역은 통계 산정(N1, N2…) 섹션도 가지며, 화면의 수치 하나마다 무엇을 세는지와 출처를 한 줄로 남깁니다 — 업무 정책과 같은 영속성·처리 방식을 따릅니다. 피처의 `summary.md`는 `Area:` 줄로 제품 영역을 적고, spec은 어떤 정책이나 수치를 유지하거나 바꾸는지 밝히고, Side effects 검증자는 건드리지 않아야 할 것이 그대로인지 다시 확인하며, close 병합이 새로 생기거나 바뀐 정책·수치를 페이지에 올립니다. `tools/kb.sh show "학습 > 평가 > 제출"`(또는 페이지 파일 이름)은 그 페이지와 그 제품 영역을 바꾼 피처를 함께 보여줍니다.
+**지식은 제품 영역별로 정리됩니다.** 웹앱이면 제품 영역은 메뉴 하나이고 메뉴 경로(`학습 > 평가 > 제출`)로 부릅니다. 다른 소프트웨어는 모듈, API, 배치 작업, CLI 명령입니다. 제품 영역마다 페이지가 한 장 있고, 파일 이름은 메뉴 경로입니다 — `memory/areas/학습 - 평가 - 제출.md`(` > `는 ` - `로, `/ \ : * ? " < > |`는 `-`로 바꿉니다). 페이지는 읽는 사람 우선입니다. 위에는 개발자가 아니어도 읽을 수 있는 문장으로 쓴 업무 정책(P1, P2…), 동작 방식, 그 영역을 바꾼 피처별 이력 한 행이 오고(정책·수치·이력·근거는 한 줄에 하나씩 쓰는 마크다운 표), 출처·검증·코드 위치는 모두 맨 아래 접힌 근거 블록에 둡니다. 건수·비율·점수·차트를 보여주는 영역은 통계 산정(N1, N2…) 섹션도 가지며, 화면의 수치 하나마다 무엇을 세는지와 출처를 한 행으로 남깁니다 — 업무 정책과 같은 영속성·처리 방식을 따릅니다. 피처의 `summary.md`는 `Area:` 줄로 제품 영역을 적고, spec은 어떤 정책이나 수치를 유지하거나 바꾸는지 밝히고, Side effects 검증자는 건드리지 않아야 할 것이 그대로인지 다시 확인하며, close 병합이 새로 생기거나 바뀐 정책·수치를 페이지에 올립니다. `tools/kb.sh show "학습 > 평가 > 제출"`(또는 페이지 파일 이름)은 그 페이지와 그 제품 영역을 바꾼 피처를 함께 보여줍니다.
 
 **기록을 다시 읽는 도구는 `tools/kb.sh`입니다.** `index`는 목차 페이지를 다시 만듭니다(`init.sh`와 `close.sh`가 자동으로 실행합니다). 페이지는 정책 수·마지막 변경·피처를 담은 제품 영역 표, 상태·날짜·제품 영역·태그를 담은 개요 표(최신순), 아직 close가 병합하지 않은 harvest 목록, 피처별 절 순서입니다. `show <slug>`는 피처 하나를 요약본으로 보여줍니다. 목표, `summary.md`(제품 영역·무엇이 문제였나·Before → After·확인 방법을 담는, 계속 갱신하도록 만든 유일한 기록), 배포 상태, 병합되지 않은 harvest 후보, 교훈 제목이 먼저 나오고 파일 경로는 마지막입니다. `search "<문자열>"`은 열린 피처와 닫힌 피처, 지속 메모리를 대상으로 출력량을 제한한 문자열 검색을 하고, `harvest [--stale <일수>]`는 harvest.md가 아직 memory/에 들어가지 않은 열린 피처를 유휴 기간과 함께 나열합니다(유휴 상태가 오래된 피처는 close 없이 병합할 수 있습니다 — AGENTS.md 규칙 4). `--area <폴더>`를 붙이면 그 폴더 안의 모든 저장소를 대상으로 같은 일을 하며, 원래 체크아웃이 사라진 피처도 읽을 수 있습니다. 저장소 config.md에 `index_style: obsidian`을 적으면 Obsidian 볼트용 frontmatter와 인라인 `#태그`를 덧붙입니다. 생성 시각은 절대 쓰지 않으므로 내용이 같으면 diff도 생기지 않습니다. 종료 코드는 `0` 찾음, `1` 없음, `2` 사용법 오류 또는 거부입니다.
 
diff --git a/README.md b/README.md
index 5a41802..7f358e1 100644
--- a/README.md
+++ b/README.md
@@ -169,7 +169,7 @@ Per feature, inside the **target project**:
 
 **Where the records live is your choice.** By default they sit in the project's working copy. `init.sh . --area ~/knowledge` puts them in a folder you choose instead — `<area>/<unit>-<checkout-id>/`, with `.sdlc` linked to it, one store per checkout so two worktrees never share approvals. The area is refused if it sits inside the project (or the project inside it), if another checkout already owns that store, if a real `.sdlc` directory is already there (nothing is ever relocated for you), or if the link cannot be made. That ownership is re-checked at RUNTIME, not only at init: `check-gate.sh`, `approve.sh`, `close.sh`, `status.sh`, `tools/auto.sh`, `tools/verify.sh` and `tools/handoff.sh` refuse before any verdict or write when `<store>/PROJECT` names a different checkout, so a copied working copy (`cp -R`, rsync and most restores keep the symlink) can neither open another checkout's gate nor close its features. Reading is never bound that way: `tools/kb.sh show|search|list` still works, and nothing is ever re-bound or moved for you. Whichever you choose, **the store is yours to back up** — git no longer does it for you.
 
-**Knowledge is filed by product area.** For a web app an area is one menu, named by its menu path (`학습 > 평가 > 제출`); for other software a module, API, job, or CLI command. Each area has one page, named by its menu path — `memory/areas/학습 - 평가 - 제출.md` (each ` > ` written ` - `; any of `/ \ : * ? " < > |` becomes `-`). The page reads reader first: its business rules numbered P1, P2… in sentences a non-developer can read, how it works, and a history line per feature that changed it; every source, verification, and code location sits in a folded evidence block at the bottom. An area that shows counts, rates, scores or charts also gets a Numbers section, N1, N2…, one line per figure recording what it counts and its source — the same permanence and treatment as a business rule. A feature's `summary.md` names its area on an `Area:` line, the spec states which rules or figures it keeps or changes, the Side effects verifier re-checks the ones it should not have touched, and the close merge files new or changed rules and figures on the page. `tools/kb.sh show "학습 > 평가 > 제출"` (or the page's file name) prints the page with the features that changed it.
+**Knowledge is filed by product area.** For a web app an area is one menu, named by its menu path (`학습 > 평가 > 제출`); for other software a module, API, job, or CLI command. Each area has one page, named by its menu path — `memory/areas/학습 - 평가 - 제출.md` (each ` > ` written ` - `; any of `/ \ : * ? " < > |` becomes `-`). The page reads reader first: its business rules numbered P1, P2… in sentences a non-developer can read, how it works, and a history row per feature that changed it — rules, figures, history and evidence are markdown tables, one row each; every source, verification, and code location sits in a folded evidence block at the bottom. An area that shows counts, rates, scores or charts also gets a Numbers section, N1, N2…, one row per figure recording what it counts and its source — the same permanence and treatment as a business rule. A feature's `summary.md` names its area on an `Area:` line, the spec states which rules or figures it keeps or changes, the Side effects verifier re-checks the ones it should not have touched, and the close merge files new or changed rules and figures on the page. `tools/kb.sh show "학습 > 평가 > 제출"` (or the page's file name) prints the page with the features that changed it.
 
 **Reading the records back** is `tools/kb.sh`: `index` regenerates the contents page (`init.sh` and `close.sh` do it for you) — the product areas with their rule count, last change and features, an overview table by state, date, area and tags, newest first, the harvests no close has merged yet, then one section per feature; `show <slug>` prints one feature as a digest — goal, its `summary.md` (area, what was wrong, before → after, how to check, the one record meant to be kept current), delivery, unmerged harvest candidates, lesson titles, then the paths; `search "<text>"` does a bounded literal search over open and closed features plus durable memory; `harvest [--stale <days>]` lists open features whose harvest.md is not in memory yet, with idle time (a stale one may be merged without closing — AGENTS.md rule 4); and `--area <folder>` does any of these across every store in that folder — including features whose checkout no longer exists. `index_style: obsidian` in the store's config.md adds frontmatter and inline `#tags` for a vault; no timestamp is ever written, so an unchanged page produces no diff. Exit codes: `0` found, `1` nothing found, `2` usage error or refusal.
 
diff --git a/gates/selftest.sh b/gates/selftest.sh
index d15e57e..16e60da 100755
--- a/gates/selftest.sh
+++ b/gates/selftest.sh
@@ -67,33 +67,36 @@ LC_ALL=en_US.UTF-8 bash "$kit/tools/kb.sh" index >/dev/null
 grep -qF '| [명단 > 내보내기](memory/areas/roster.md) | 1 | — | f1 |' .sdlc/README.md || fail "area table wrong: $(grep '명단' .sdlc/README.md)"
 out=$(bash "$kit/tools/kb.sh" show "명단 > 내보내기")
 case "$out" in (*"P1: 현재 학기만"*) ;; (*) fail "show <menu path> did not print the rule: $out";; esac
-#    a page named by its menu path (" > " → " - "), reader first: plain rules on
-#    top, evidence in <details> at the bottom — evidence lines do not count as rules
+#    a page named by its menu path (" > " → " - "), reader first, in tables:
+#    rules counted from rows above the evidence block (evidence and retired rows
+#    are not), last change from the History table, header words free
 mkdir -p .sdlc/work/f3; printf -- '- Area: 교사 > 학생 > 학급 분석\n' > .sdlc/work/f3/summary.md
-printf -- '%s\n' '# Area: 교사 > 학생 > 학급 분석' '- Menu: 교사 > 학생 > 학급 분석' '## Business rules (정책)' \
-  '- P1: 자기 학급만 본다' '- P2: 전학생은 빠진다' '- ~~P3: 지난 학기도 보인다~~' '## History' '- 2026-09-01 f3 — 참여율 추가' \
-  '<details>' '<summary>근거 · 코드 위치 (개발자용)</summary>' '' '- Where: ClassAnalysis#get' \
-  '- P1 — source: 기획서 · set by f3' '- P3 — retired 2026-09-01 by f3: 정책 변경' '' '</details>' > ".sdlc/memory/areas/교사 - 학생 - 학급 분석.md"
+printf -- '%s\n' '# Area: 교사 > 학생 > 학급 분석' '- Menu: 교사 > 학생 > 학급 분석' '## Business rules (정책)' '| # | 정책 |' '|---|---|' \
+  '| P1 | 자기 학급만 본다 |' '| P2 | 전학생은 빠진다 |' '| ~~P3~~ | ~~지난 학기도 보인다~~ |' '| P4 | <rule> |' \
+  '## History' '| 날짜 | 작업 | 바뀐 점 |' '|---|---|---|' '| 2026-09-01 | f3 | 참여율 추가 |' '| 2026-08-01 | f1 | 처음 |' \
+  '<details>' '<summary>근거 · 코드 위치 (개발자용)</summary>' '' '- Where: ClassAnalysis#get' '' '| # | 출처 | 작업 | 검증 |' '|---|---|---|---|' \
+  '| P1 | 기획서 | f3 | code — 2026-09-01 |' '| P3 | retired 2026-09-01 by f3: 정책 변경 | f3 | human |' '' '</details>' > ".sdlc/memory/areas/교사 - 학생 - 학급 분석.md"
+#    the Obsidian form of the evidence block: a folded callout, lines prefixed "> "
+printf -- '%s\n' '- Menu: 학생 > 과제' '## Business rules (정책)' '| # | Rule |' '|---|---|' '| P1 | 마감 후 제출 불가 |' \
+  '## History' '| Date | Feature | What changed |' '|---|---|---|' '| 2026-09-02 | f4 | 마감 표시 |' \
+  '> [!info]- 근거 · 코드 위치 (개발자용)' '> - Where: HomeworkApi#submit' '>' '> | # | Source | Set by | Verified |' '> |---|---|---|---|' \
+  '> | P1 | 기획서 | f4 | test — 2026-09-02 |' '> | P2 | 증거 행은 규칙이 아니다 | f4 | — |' > ".sdlc/memory/areas/학생 - 과제.md"
 LC_ALL=en_US.UTF-8 bash "$kit/tools/kb.sh" index >/dev/null
 grep -qF '| [교사 > 학생 > 학급 분석](memory/areas/교사%20-%20학생%20-%20학급%20분석.md) | 2 | 2026-09-01 | f3 |' .sdlc/README.md \
-  || fail "menu-named page row wrong: $(grep '교사' .sdlc/README.md)"
+  || fail "menu-named table page row wrong: $(grep '교사' .sdlc/README.md)"
+grep -qF '| [학생 > 과제](memory/areas/학생%20-%20과제.md) | 1 | 2026-09-02 | — |' .sdlc/README.md || fail "callout page row wrong: $(grep '과제' .sdlc/README.md)"
 for q in "교사 > 학생 > 학급 분석" "교사 - 학생 - 학급 분석"; do
   out=$(bash "$kit/tools/kb.sh" show "$q") || fail "show '$q' found no page"
-  case "$out" in (*"P1: 자기 학급만"*"근거 · 코드 위치 (개발자용):"*"Where: ClassAnalysis#get"*) ;; (*) fail "show '$q' not reader first: $out";; esac
+  case "$out" in (*"| P1 | 자기 학급만"*"| 2026-09-01 | f3 |"*"근거 · 코드 위치 (개발자용):"*"Where: ClassAnalysis#get"*"| P3 | retired"*) ;; (*) fail "show '$q' not reader first: $out";; esac
+  case "$out" in (*"<rule>"*) fail "show printed a placeholder row: $out";; esac
 done
-bash "$kit/tools/kb.sh" show "../areas/roster" >/dev/null 2>&1 && fail "area lookup followed a path"
-#    the Obsidian form of the evidence block: a folded callout, lines prefixed "> "
-printf -- '%s\n' '- Menu: 학생 > 과제' '## Business rules (정책)' '- P1: 마감 후 제출 불가' '## History' '- 2026-09-02 f4 — 마감 표시' \
-  '> [!info]- 근거 · 코드 위치 (개발자용)' '> - Where: HomeworkApi#submit' '> - P1 — source: 기획서 · set by f4' '> - P2: 콜론 줄도 규칙이 아니다' \
-  > ".sdlc/memory/areas/학생 - 과제.md"
-LC_ALL=en_US.UTF-8 bash "$kit/tools/kb.sh" index >/dev/null
-grep -qF '| [학생 > 과제](memory/areas/학생%20-%20과제.md) | 1 | 2026-09-02 | — |' .sdlc/README.md || fail "callout page row wrong: $(grep '과제' .sdlc/README.md)"
 out=$(bash "$kit/tools/kb.sh" show "학생 > 과제") || fail "show found no callout page"
-case "$out" in (*"P1: 마감 후"*"History:"*"근거 · 코드 위치 (개발자용):"*"  - Where: HomeworkApi#submit"*) ;; (*) fail "callout evidence not read last: $out";; esac
-case "$out" in (*"> -"*) fail "callout prefix printed: $out";; esac
+case "$out" in (*"| P1 | 마감 후"*"History:"*"근거 · 코드 위치 (개발자용):"*"  - Where: HomeworkApi#submit"*"  | P1 | 기획서 | f4 |"*) ;; (*) fail "callout evidence not read last: $out";; esac
+printf '%s\n' "$out" | grep -q '^ *>' && fail "callout prefix printed: $out"
 fns=$(sed -n -e '/^kb_field() {/,/^}/p' -e '/^kb_get() {/,/^}/p' "$kit/tools/kb.sh")
 where=$(eval "$fns"; kb_get ".sdlc/memory/areas/학생 - 과제.md" Where)
 [ "$where" = "HomeworkApi#submit" ] || fail "kb_get did not read Where inside the callout: '$where'"
+bash "$kit/tools/kb.sh" show "../areas/roster" >/dev/null 2>&1 && fail "area lookup followed a path"
 echo "ok: knowledge by product area"
 
 echo "SELFTEST PASS"
diff --git a/log/changelog-v0.16.0.md b/log/changelog-v0.16.0.md
index d8eebaa..4d09165 100644
--- a/log/changelog-v0.16.0.md
+++ b/log/changelog-v0.16.0.md
@@ -9,13 +9,14 @@ filing rules as a business rule. It also names each area page by its menu
 path, so the file list reads like the product's menus, and lays the page out
 reader first: plain language on top for a non-developer, every source,
 verification, and code location folded into an evidence block at the
-bottom. Gates, approvals, and the `--json` schema are unchanged; `kb.sh`'s
-rule count still counts only the live `- P<n>:` rules.
+bottom, with rules, figures, history and evidence as markdown tables. Gates,
+approvals, and the `--json` schema are unchanged; `kb.sh`'s rule count still
+counts only live rules (a `| P<n> |` row, or an older `- P<n>:` line).
 
 ## Changes
 
 - **`templates/area.md`** gains an optional **Numbers (통계 산정)** section
-  between Business rules and How it works: one line per on-screen figure,
+  between Business rules and How it works: one row per on-screen figure,
   named as the screen labels it, recording what it counts, out of what,
   whether it is real-time or batch (job, refresh, data-as-of), its source,
   and who set it. Delete the section for an area with no figures. N-numbers
@@ -52,17 +53,18 @@ rule count still counts only the live `- P<n>:` rules.
   prints a `<summary>…</summary>` line as a section label, so `show` prints
   the page top first and the evidence block last, under
   `근거 · 코드 위치 (개발자용):` — the part the line bound cuts first. The
-  Rules column no longer counts an unfilled template line `- P1: <…>`,
-  and `show` drops such a line (`- <label with digits>: <…>`) the way it
-  already dropped `- Where: <…>`.
+  Rules column no longer counts an unfilled template row `| P1 | <…> |`
+  (or line `- P1: <…>`), and `show` drops such a line the way it already
+  dropped `- Where: <…>`.
 - **Reader-first `templates/area.md`.** Menu and Aliases on top; Business
   rules, Numbers, How it works, and History in plain sentences with no
   source and no code; then a `<details>` block — "근거 · 코드 위치
-  (개발자용)" — holding the Where line and one evidence line per rule or
-  figure: `- P1 — source: … · set by <slug> · [verified: how — date]`. The
-  evidence line has no colon after the number, so it is never counted as a
-  rule. A retired rule stays struck through on top (`- ~~P3: …~~`) and its
-  evidence line carries `retired YYYY-MM-DD by <slug>: <why>`. The section
+  (개발자용)" — holding the Where line and one evidence row per rule or
+  figure: `| P1 | source | set by | verified |`. Evidence rows sit below
+  the block's opening line, where the rule count stops, so they are never
+  counted as rules. A retired rule stays struck through on top
+  (`| ~~P3~~ | ~~rule~~ |`) and its evidence row's Source cell carries
+  `retired YYYY-MM-DD by <slug>: <why>`. The section
   headings are unchanged. `templates/harvest.md`'s area candidates split
   the plain sentence from its evidence the same way, and ship's
   retrospective says which part goes where.
@@ -71,6 +73,20 @@ rule count still counts only the live `- P<n>:` rules.
   a folded callout (`> [!info]- 근거 · 코드 위치 (개발자용)`, every line
   prefixed `> `); `kb.sh` reads both forms (Where found, callout lines
   printed last without `> `, never counted as rules).
+- **Tables for rules, figures, history and evidence.** Each is one
+  markdown table row (`| P1 | <rule> |`, `| N1 | label | counted | out of
+  | timing |`, `| YYYY-MM-DD | slug | what changed |`, and in the evidence
+  block `| P1 | source | set by | verified |` under the `- Where:` line);
+  Menu, Aliases and How it works stay list lines. A retired rule is
+  `| ~~P3~~ | ~~rule~~ |`, its reason in its evidence row's Source cell.
+  Header words are free (`| # | 정책 |` works); a literal `|` in a cell is
+  `\|`. `kb.sh` keys on row shapes only: the Rules count reads `| P<n> |`
+  rows above the evidence block (the `<details>` or `> [!…]` line), the
+  last change reads History's first `| YYYY-MM-DD |` row, and `show`
+  prints tables as written — callout rows without `> ` — dropping only a
+  template placeholder row (every cell after the first `<…>`) and the
+  header of a table left with no rows. The close merge (AGENTS.md rule 4,
+  ship, harvest.md) writes rows; harvest candidates stay line-shaped.
 - **README.md / README.ko.md**: the "Knowledge is filed by product area"
   paragraph and the memory-tree comment now mention the optional Numbers
   section, the menu-path file name, and the folded evidence block.
@@ -92,6 +108,11 @@ tools/kb.sh index
 instead of the Menu should be changed to the Menu words. Nothing is renamed
 for you.
 
+**List-form pages keep working.** A page written with `- P<n>: rule`
+lines and `- YYYY-MM-DD <slug> — …` History lines (v0.15.0 and the first
+v0.16.0 commits) is still counted and dated, and a page may mix both forms;
+convert it to rows whenever the close merge next edits it.
+
 **Existing pages with inline sources keep working.** A rule written the old
 way — `- P1: <rule> — source: … · set by …` — still counts and still
 prints; the close merge moves its source to the evidence block the next
@@ -107,9 +128,13 @@ with only P-numbers renders identically to before). No script parses the
 
 - `bash gates/selftest.sh` → `SELFTEST PASS`. Its "knowledge by product
   area" test now also files a feature under a Hangul menu page named
-  `교사 - 학생 - 학급 분석.md`, laid out reader first with evidence lines and
+  `교사 - 학생 - 학급 분석.md`, laid out reader first with evidence rows and
   a retired rule: the contents-page row links the percent-encoded path with
-  a Rules count of 2 (evidence and retired lines not counted); `show` by
+  a Rules count of 2 (evidence and retired rows not counted); `show` by
   Menu and by file name prints the rules before the evidence block; a
   lookup holding `/` is refused; a second page in the Obsidian callout form
-  is read the same way. With the old `kb.sh`, the test fails.
+  is read the same way. Both pages are in table form with Korean header
+  words: evidence rows, a retired row and a placeholder row are not
+  counted, the last change comes from the History table, and `show`
+  prints the tables with the evidence last. The list-form `roster` page
+  still counts. With the old `kb.sh`, the test fails.
diff --git a/skills/5-ship/SKILL.md b/skills/5-ship/SKILL.md
index 0f3899d..f136cd5 100644
--- a/skills/5-ship/SKILL.md
+++ b/skills/5-ship/SKILL.md
@@ -72,9 +72,9 @@ evidence, the source (the origin, the spec R-item or compact intent O-item)
 and how it was verified — plus one history line per area it changed. **A
 figure the area page shows a Numbers section for is treated the same way**:
 the N-number it sets, changes, or retires, named on screen, with what it
-counts, and its source as evidence. The close merge writes the sentence on
-top of the area page and the evidence on the page's evidence line in the
-`<details>` block at the bottom (or the folded callout in an Obsidian store,
+counts, and its source as evidence. The close merge writes the sentence as
+a table row on top of the area page (a history candidate as a History row)
+and the evidence as its evidence row in the `<details>` block at the bottom (or the folded callout in an Obsidian store,
 index_style: obsidian); a page it creates is named by its menu path — the Menu with each
 ` > ` written ` - ` and any of `/ \ : * ? " < > |` replaced with `-`
 (templates/area.md). Then finish the feature's `summary.md`:
diff --git a/templates/area.md b/templates/area.md
index 7ebce2c..ec9cee5 100644
--- a/templates/area.md
+++ b/templates/area.md
@@ -7,10 +7,13 @@
      For a web app an area is one menu — split a page when its sub-screens
      carry different rules or it passes ~60 lines. Other software: a module,
      API, job, or CLI command. ONE writer: the close merge (AGENTS.md rule 4).
-     READER FIRST: everything above <details> is for a non-developer — plain
-     sentences, no source, no code. Every source, date, verification, and code
-     identifier goes in the <details> block at the bottom, one evidence line
-     per rule or figure. -->
+     READER FIRST: everything above the evidence block is for a non-developer
+     — plain sentences, no source, no code. Every source, date, verification,
+     and code identifier goes in the evidence block at the bottom, one row per
+     rule or figure.
+     TABLES: rules, figures, history and evidence are one table row each. The
+     header words are free (a store may write | # | 정책 |); kb.sh reads only
+     the row shapes. Write a literal "|" inside a cell as "\|". -->
 
 - Menu: <menu path as users see it; no commas>
 - Aliases: <other words people search with>
@@ -19,46 +22,58 @@
      Aliases: screen title, ticket wording, other language. -->
 
 ## Business rules (정책)
-<!-- The product's rules for this area, one plain sentence each, testable,
-     nothing else on the line. Its source goes on its evidence line below.
-     P-numbers are permanent. The close merge gives a new rule the next number
-     this page has never used (retired ones count), edits a changed rule in
-     place (and its evidence line), and keeps a retired one struck through:
-       - ~~P3: <rule>~~
-     with the reason on its evidence line:
-       - P3 — retired YYYY-MM-DD by <slug>: <why> · source: <…>
+<!-- One row per rule, one plain testable sentence, nothing else in the row.
+     Its source goes in its evidence row below. P-numbers are permanent: the
+     close merge gives a new rule the next number this page has never used
+     (retired ones count), edits a changed rule in place (and its evidence
+     row), and keeps a retired one struck through:
+       | ~~P3~~ | ~~<rule>~~ |
+     with the reason in its evidence row's Source cell:
+       | P3 | retired YYYY-MM-DD by <slug>: <why> | … | … |
      Not POLICY.md — that file holds the human's rules for agents. -->
-- P1: <rule in one plain sentence — no source, no code>
+| # | Rule |
+|---|---|
+| P1 | <rule in one plain sentence — no source, no code> |
 
 ## Numbers (통계 산정)
 <!-- Only for an area that shows counts, rates, scores or charts; delete the
-     section otherwise. One line per figure, named as the screen labels it.
+     section otherwise. One row per figure, named as the screen labels it.
      N-numbers follow the P-number rules above: permanent, a changed figure is
-     edited in place, a retired one is kept struck through with the reason on
-     its evidence line. The batch job's name goes on the evidence line. -->
-- N1: <label on screen> — <what is counted> / <out of what> · <real-time | batch: refreshed <when>, data up to <when>>
+     edited in place, a retired one is kept struck through with the reason in
+     its evidence row. The batch job's name goes in the evidence row. -->
+| # | Label on screen | What is counted | Out of | Timing |
+|---|---|---|---|---|
+| N1 | <label> | <what is counted> | <out of what, or —> | <real-time \| batch: refreshed <when>, data up to <when>> |
 
 ## How it works
-<!-- What the user does, step by step, what the system does in response, and
-     facts that hold for this area only. Use DOMAIN.md terms; do not redefine
-     them here. -->
-<≤10 plain lines>
+<!-- ≤10 plain bullets: what the user does, step by step, what the system does
+     in response, and facts that hold for this area only. Use DOMAIN.md terms;
+     do not redefine them here. -->
+- <step or fact, in plain words>
 
 ## History
-<!-- Newest first; one line per shipped feature that changed this area, from
+<!-- Newest first; one row per shipped feature that changed this area, from
      its summary.md. -->
-- YYYY-MM-DD <slug> — <what changed for the user>
+| Date | Feature | What changed for the user |
+|---|---|---|
+| YYYY-MM-DD | <slug> | <…> |
 
 <!-- In a store with `index_style: obsidian` write this block as a folded
      callout instead, every line prefixed "> ":
        > [!info]- 근거 · 코드 위치 (개발자용)
        > - Where: <…>
-       > - P1 — source: … · set by … · [verified: …] -->
+       >
+       > | # | Source | Set by | Verified |
+       > |---|---|---|---|
+       > | P1 | … | … | … | -->
 <details>
 <summary>근거 · 코드 위치 (개발자용)</summary>
 
 - Where: <route · file:symbol · endpoint · job>
-- P1 — source: <기획서 / ticket / human / code, date> · set by <slug> · [verified: how — YYYY-MM-DD]
-- N1 — source: … · set by … · [verified: …]
+
+| # | Source | Set by | Verified |
+|---|---|---|---|
+| P1 | <기획서 / ticket / human / code, date> | <slug> | <how — YYYY-MM-DD> |
+| N1 | <…> | <slug> | <how — YYYY-MM-DD> |
 
 </details>
diff --git a/templates/harvest.md b/templates/harvest.md
index 5966d40..9d12024 100644
--- a/templates/harvest.md
+++ b/templates/harvest.md
@@ -12,8 +12,9 @@
      - [area: <menu path>] history: YYYY-MM-DD <slug> — <what changed for the user>
      Rules, figures, and history merge only when the feature closes shipped;
      a stale merge leaves them here (rule 4). The merge numbers a new rule
-     or figure, writes the plain sentence on top of the area page and the
-     evidence part on its evidence line in the <details> block at the bottom,
+     or figure, writes the plain sentence as a table row on top of the area
+     page (history as a History row) and the evidence part as its evidence
+     row in the <details> block at the bottom,
      or the folded callout in an Obsidian store (index_style: obsidian;
      templates/area.md).
      Code identifiers belong only in the evidence part. -->
diff --git a/tools/kb.sh b/tools/kb.sh
index 1968c28..5a751cd 100755
--- a/tools/kb.sh
+++ b/tools/kb.sh
@@ -147,11 +147,19 @@ kb_hashtags() { # "tag, tag" → "#tag #tag" for Obsidian's tag pane (spaces ins
 # Obsidian form) become `X:` labels — the callout's own lines print without
 # their "> " prefix — so a record embedded in the
 # contents page cannot hijack its outline. A heading with nothing under it yet
-# is dropped too: an unfinished record prints only what is known.
+# is dropped too: an unfinished record prints only what is known. A markdown
+# table prints as written, except a template placeholder row (every cell after
+# the first is only `<…>`); a table left with no data row loses its header too.
 kb_body() { # <file>
   [ -f "$1" ] || return 0
   awk '
-    BEGIN { c = 0; q = 0 }
+    function ph(l,   a, n, i, x) { # a template placeholder row: every cell after the first is <…>
+      gsub(/\\\|/, "\001", l); sub(/^[ \t]*\|/, "", l); sub(/\|[ \t]*$/, "", l)
+      n = split(l, a, "|")
+      for (i = 2; i <= n; i++) { x = a[i]; gsub(/^[ \t]+|[ \t]+$/, "", x); if (x !~ /^<.*>$/) return 0 }
+      return n >= 2
+    }
+    BEGIN { c = 0; q = 0; tb = 0 }
     {
       line = $0
       if (c) { i = index(line, "-->"); if (i) { c = 0; line = substr(line, i + 3) } else next }
@@ -166,6 +174,15 @@ kb_body() { # <file>
       if (line ~ /^# /) next
       if (line ~ /^##+ /) { sub(/^#+ +/, "", line); pend = line ":"; next }
       if (line ~ /^[ \t]*<summary>.*<\/summary>[ \t]*$/) { sub(/^[ \t]*<summary>[ \t]*/, "", line); sub(/[ \t]*<\/summary>[ \t]*$/, "", line); pend = line ":"; next }
+      if (line ~ /^[ \t]*\|/) {   # a table: hold the header and separator until a data row shows
+        if (!tb) { tb = 1; hd = line; hs = ""; next }
+        if (hd != "" && hs == "" && line ~ /^[ \t]*\|[ \t:|-]+$/) { hs = line; next }
+        if (ph(line)) next
+        if (pend != "") { print pend; pend = "" }
+        if (hd != "") { print hd; if (hs != "") print hs; hd = ""; hs = "" }
+        print line; next
+      }
+      tb = 0; hd = ""; hs = ""
       if (line ~ /^[ \t]*$/) next
       if (line ~ /^[ \t]*[-*] *<[^>]*>[ \t]*$/) next
       if (line ~ /^[ \t]*<[^>]*>[ \t]*$/) next
@@ -225,13 +242,23 @@ kb_area_menu() { # <page> → its Menu line, else its H1, else its file name (an
   case "$m" in ''|'<'*) m=$(basename "$1" .md);; esac
   printf '%s' "$m"
 }
-# Live rules: the top "- P<n>: rule" lines. A retired one reads "- ~~P…", an
-# evidence line "- P<n> — source…" (no colon after the number; "> - P<n> —" in
-# an Obsidian callout — never at column 0), and an unfilled
-# template line "- P1: <…>" — none of them counts.
-kb_area_rules() { awk '/^- P[0-9]+:/ && !/^- P[0-9]+: *</ { n++ } END { print n + 0 }' "$1"; }
-kb_area_last() { # <page> → the date of the newest History line
-  awk '/^## /{ h = ($0 ~ /^## History/); next } h && /^- [0-9][0-9][0-9][0-9]-[0-9][0-9]-[0-9][0-9]/ { print substr($0, 3, 10); exit }' "$1"
+# Live rules: the "| P<n> | rule |" table rows, and the "- P<n>: rule" list
+# lines pages written before the table form still hold — counted only ABOVE
+# the evidence block (the `<details>` or `> [!…]` line), whose rows carry the
+# same P-numbers. Row shapes only, never header words. A retired rule reads
+# "| ~~P…" or "- ~~P…", and an unfilled template row "| P1 | <…> |" (or line
+# "- P1: <…>") — none of them counts.
+kb_area_rules() {
+  awk '/^[ \t]*<details/ || /^> *\[!/ { exit }
+    /^- P[0-9]+:/ && !/^- P[0-9]+: *</ { n++; next }
+    /^\| *P[0-9]+ *\|/ { r = $0; sub(/^\| *P[0-9]+ *\| */, "", r); if (r !~ /^</) n++ }
+    END { print n + 0 }' "$1"
+}
+kb_area_last() { # <page> → the date of History's first (newest) row "| YYYY-MM-DD |", or list line "- YYYY-MM-DD"
+  awk '/^[ \t]*<details/ || /^> *\[!/ { exit }
+    /^## /{ h = ($0 ~ /^## History/); next }
+    h && (/^- [0-9][0-9][0-9][0-9]-[0-9][0-9]-[0-9][0-9]/ || /^\| *[0-9][0-9][0-9][0-9]-[0-9][0-9]-[0-9][0-9] *\|/) {
+      match($0, /[0-9][0-9][0-9][0-9]-[0-9][0-9]-[0-9][0-9]/); print substr($0, RSTART, 10); exit }' "$1"
 }
 # "<area><SEP><kind><SEP><slug><SEP><dir>" for every feature naming an area.
 kb_area_refs() { # <store>