### Example Frontmatter for Image Ingest

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/ingest/references/images.md

YAML frontmatter example for an image input, including title, type, tags, update date, summary, source, media format, resolution, and capture date.

```yaml
---
title: K-POP 비주얼 디렉션 무드보드
type: inbox
tags: [비주얼디렉션, 무드보드, K-POP]
updated: 2026-04-21
summary: 파스텔 톤 인물 클로즈업을 3×3 그리드로 배치한 K-POP 무드보드. 채도 낮춤 + 필름 질감으로 레트로 분위기 강조.
source:
media: image
format: png
resolution: 2048x1152
captured: 2026-04-19
---
```

--------------------------------

### Parse sips Output for Metadata

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/ingest/references/images.md

Example of `sips` command output and how to interpret it to extract relevant metadata fields like pixel dimensions and format.

```text
/path/to/file.png
  pixelWidth: 1920
  pixelHeight: 1080
  format: png
```

--------------------------------

### Frontmatter Schema Validation: Warn Example

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/lint/references/checks.md

Example of a frontmatter schema warning for summary length. The summary is outside the recommended character count, suggesting expansion.

```text
- notes/디자인-사고.md: `summary` 길이 위반 [severity: warn]
  현재: 32자
  기대: 80~120자
  제안: 핵심 논점 한 줄 추가로 확장
```

--------------------------------

### Plugin Setup and Skill Addition

Source: https://context7.com/choi-sanadesu/memography/llms.txt

Instructions for setting up Memography plugins in local development mode and validating YAML files. Each skill is defined by a single SKILL.md file, with the 'description:' field in the frontmatter used by Claude for automatic triggering.

```bash
# Install plugin (local development mode before registry registration)
ln -s /path/to/memography ~/.claude/plugins/memography

# Validate plugin
claude plugin validate

# Batch validate YAML
python3 -c "import glob, yaml; [list(yaml.safe_load_all(open(f))) for f in glob.glob('**/*.y*ml', recursive=True, include_hidden=True) if '/.git/' not in f and '/node_modules/' not in f]"
```

--------------------------------

### Automated Processing Report Example

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/classify/references/workflow.md

Example of the automated processing report showing changes to entity files.

```text
[Auto] entities/people/{name}.md
  + related.sources: "[[sources/2026-04-23-newjeans-comeback]]"

[Auto] entities/brands/{label}.md
  + related.sources: "[[sources/2026-04-23-pdf-report]]"
  + 본문 활동 기록: "2026-04: {사건} ([[source]])"
```

--------------------------------

### Inbox Residue Report Example

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/lint/references/checks.md

Example report for the inbox residue check, showing the number of remaining files, their details, and any files exceeding the warning threshold for being stale.

```text
잔여: {N}개
파일 목록:
- {filename} (크기: {size}, 수정: {mtime}, {경과일}일 경과)

⚠️ {threshold} 초과 항목:
- {filename} ({경과일}일)

⚠️ 총 {N}개 — 임계 10개 초과
```

--------------------------------

### Orphaned Page Report Example

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/lint/references/checks.md

Example report item for an orphaned page, indicating it has zero backlinks. It includes path, backlink count, type, and update date, with a hint for further investigation.

```text
- {path} (backlinks: 0)
  type: {type}
  updated: {date}
  판단 힌트: 최근 추가됐는지, 다른 페이지에서 언급되는데 링크가 빠졌는지 등
```

--------------------------------

### Example Vault Index.md Structure

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/classify/references/workflow.md

Illustrative structure of a Vault's index.md file, showing how different entity types are organized into sections.

```markdown
## entities
- [[entities/people/뉴진스|뉴진스]] — ...

## sources
- [[sources/2026-04-23-newjeans-comeback|NewJeans 컴백]] — ...

## cases
- ...
```

--------------------------------

### Install Local Development Mode Plugin

Source: https://github.com/choi-sanadesu/memography/blob/main/README.md

Add the plugin path to your Claude Code settings or create a symbolic link in ~/.claude/plugins/ for local development before the plugin is registered in the plugin registry.

```bash
# Claude Code 설정에서 플러그인 경로 추가
# 또는 ~/.claude/plugins/ 심볼릭 링크
```

--------------------------------

### Memography Lint Report Example

Source: https://context7.com/choi-sanadesu/memography/llms.txt

Example output of the Memography lint command, detailing vault health check results. It includes a summary table of issues found across different categories like frontmatter schema, broken links, and tag catalog violations, along with specific details for each failure.

```text
# 📚 Memography Lint Report
#
# Date: 2025-04-28
# Vault: /Users/sana/vault
# Files Scanned: 142
#
# ---
# ## Summary
#
# | # | Check                  | Status | Issues |
# |---|-------------------|---|---|
# | 1 | frontmatter schema    | ❌   | 2       |
# | 2 | Broken wiki links          | ❌   | 1       |
# | 3 | Orphaned pages            | ℹ️   | 1       |
# | 4 | Inbox residue            | ⚠️   | 4       |
# | 5 | stale updated:       | ✅   | 0       |
# | 6 | Tag catalog violation     | ⚠️   | 1       |
#
# Overall: 3 rule violations, prioritize action.
#
# ---
# ## 1. frontmatter schema
# ### ❌ Fail (2 issues)
# - `entities/people/newjeans.md`: Missing required field `summary` [severity: fail]
#   Current: (none)
#   Expected: 80-120 character one-line summary (Vault §2)
#   Suggestion: Condense the first sentence of the first H2 section
# - `sources/2026-04-spring.md`: Missing `source:` field for `type: source` page [severity: fail]
#   Current: (none)
#   Expected: Source URL (Vault §3)
#
# ## 2. Broken wiki links
# ### ❌ 1 issue
# - `cases/2026-spring-campaign.md` → `[[entities/people/erika-de-casier]]`
#   Location: related.entities
#   Status: unresolved
#   Candidates: (none)
#
# ## 4. Inbox residue
# 4 residues:
# - `inbox/2025-04-23-smerz-interview.md` (12KB, Modified: 2025-04-23, 5 days old)
# ⚠️ Priority item: 5 days old, → Recommend batch processing with `/classify`.
#
# ## 6. Tag catalog violation
# ### ⚠️ 1 issue
# - `notes/methodology.md`
#   Unregistered tags: [`methodology-theory`]
#   Suggestion: Add to `meta/tags.md` or replace with existing `methodology` tag

```

--------------------------------

### Frontmatter Schema Validation: Fail Example

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/lint/references/checks.md

Example of a frontmatter schema violation where a required field is missing. This indicates a severe violation.

```text
- entities/people/뉴진스.md: `summary` 필수 필드 누락 [severity: fail]
  현재: (없음)
  기대: 80~120자 한 줄 요약 (볼트 §2)
  제안: 첫 H2 섹션의 첫 문장을 압축
```

--------------------------------

### Fallback Metadata Extraction without sips

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/ingest/references/images.md

When `sips` is unavailable, use the `file` command to infer format and `stat` to get the creation date. Extract format and resolution from `file` output using regex if possible; otherwise, proceed with empty values.

```bash
file {path}           # 포맷 추정
stat {path}           # 생성일
```

--------------------------------

### Extract Image Metadata with sips and stat

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/ingest/references/images.md

Use `sips` to extract pixel dimensions and format, and `stat` to get the creation date and byte size of an image file. This is performed before reading the image in Step 1.

```bash
sips -g pixelWidth -g pixelHeight -g format {path}
stat -f '%SB' -t '%Y-%m-%d' {path}      # creation date
stat -f '%z' {path}                     # byte size (참고용)
```

--------------------------------

### Related Structured Link Format

Source: https://context7.com/choi-sanadesu/memography/llms.txt

Example of the YAML format used for defining structured links between different types of entities within the system.

```yaml
related:
  entities:
    - "[[entities/people/Smerz|Smerz]]"
    - "[[entities/brands/ADOR|ADOR]]"
  sources:
    - "[[sources/2025-04-28-smerz-believer-pitchfork]]"
  cases: []
  references: []
  notes: []
```

--------------------------------

### Stale Updated Field Report Example

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/lint/references/checks.md

Example report item for a page with a stale 'updated:' field. It shows the path, the outdated date with days elapsed, the page type, and a hint for review.

```text
- {path}
  updated: {date} ({N}일 경과)
  type: {type}
  판단 힌트: 내용 재검토 또는 updated 필드만 갱신
```

--------------------------------

### Example Related Structure

Source: https://github.com/choi-sanadesu/memography/blob/main/CLAUDE.md

This YAML structure defines relationships between different types of vault items. It is used for organizing and linking content within the knowledge base.

```yaml
related:
  entities: []
  sources: []
  cases: []
  references: []
  notes: []
```

--------------------------------

### Example Markdown Body for Image Ingest

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/ingest/references/images.md

Markdown structure for an image input, including the image embed, caption, OCR extracted text, and optional observation section.

```markdown
![[raw/images/moodboard-2026-04.png]]

## 캡션

파스텔 톤의 인물 클로즈업 9컷을 3×3 그리드로 배치한 무드보드. 전체적으로 채도를 낮추고 필름 질감을 얹어 레트로한 분위기를 유지한다.

## 추출 텍스트 (OCR)

> "SPRING 2026 / MOOD BOARD / REF-017"

- 언어: en

## 띠니의 관찰

…
```

--------------------------------

### Step 2 배치 뷰: 분류 제안 테이블

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/classify/references/workflow.md

3개 이상의 항목이 있을 때, 분류 제안을 테이블 형식으로 보여줍니다. 각 항목의 inbox 파일, 요약, 제안된 목적지, 신뢰도, 근거를 포함합니다.

```markdown
## R1 분류 제안 ({N}개)

| # | inbox 파일 | summary | → 목적지 | confidence | 근거 |
|---|---|---|---|---|---|
| 1 | 2026-04-23-newjeans-comeback | NewJeans 컴백... | sources/ | high | source URL + 외부 인물 |
| 2 | 2026-04-23-design-method | 디자인 사고 방법... | notes/ | high | 방법론, source 없음 |
| 3 | 2026-04-23-spring-moodboard | 스프링 무드보드... | references/ | low | media:image, 큐레이션 vs sources |
| 4 | 2026-04-23-erika-de-casier | Erika de Casier 신곡... | entities/people/ | medium | 새 엔티티 시사 (✋ 신규 생성) |
| 5 | ... | ... | ... | ... | ... |

⚠️ low confidence: #3 — references vs sources, 사나 판단 필요
⚠️ 신규 엔티티 시사: #4 — 사나 승인 후 생성
```

--------------------------------

### Step 3-3: 신규 엔티티 생성 제안 B모드

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/classify/references/workflow.md

inbox 항목이 새로운 엔티티를 정의할 때, 생성 제안을 위한 B모드 메시지 형식입니다. 엔티티 이름, 카테고리, 근거, 별칭 후보, 요약 등을 포함합니다.

```markdown
## 신규 엔티티 생성 제안

inbox/{file}.md → entities/{sub}/{name}.md

- 이름: {name}
- 카테고리: {sub} (people | brands | works | concepts | media)
- 근거: {summary 인용}
- aliases 후보: [{한글 표기}, {약칭}]
- summary: {80~120자}

생성해도 될까요?
```

--------------------------------

### Frontmatter Schema Validation: Source URL Fail Example

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/lint/references/checks.md

Example of a frontmatter schema violation where a 'source' type page is missing the required 'source:' URL field. This is a critical failure.

```text
- sources/2026-04-spring.md: `type: source` 페이지인데 `source:` 누락 [severity: fail]
  현재: (없음)
  기대: 원본 URL (볼트 §3)
```

--------------------------------

### 이미지 파일 메타데이터 추출 (Bash)

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/ingest/references/sources.md

이미지 파일의 메타데이터(픽셀 크기, 형식, 생성 날짜)를 추출하는 Bash 명령어입니다. `sips`와 `stat` 명령어를 사용합니다.

```bash
sips -g pixelWidth -g pixelHeight -g format "raw/images/ref.jpg"
stat -f '%SB' -t '%Y-%m-%d' "raw/images/ref.jpg"
```

--------------------------------

### Step 3-3: 신규 엔티티 YAML 템플릿

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/classify/references/workflow.md

새로운 엔티티 페이지 생성 시 적용되는 YAML frontmatter 템플릿입니다. title, type, tags, updated, summary, aliases, related 필드를 포함합니다.

```yaml
---
title: {정제된 이름}
type: entity
tags: [{2-3 from inbox 항목 컨텍스트}]
updated: {YYYY-MM-DD}
summary: {80-120자 — inbox 항목 컨텍스트로 합성}
aliases: [{한글 표기 등}]
related:
  entities: []
  sources: [{inbox 항목 출처면 그 링크}]
  cases: []
  references: []
  notes: []
---

# {name}

## 개요
{1-2문단 — inbox 항목 요약 기반}

## 상세
{비워둠 — 사나가 점진적으로 채움}
```

--------------------------------

### Step 3-1: 파일 이동 Bash 명령어

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/classify/references/workflow.md

Step 3에서 각 inbox 파일을 결정된 최종 목적지로 이동시키는 Bash 명령어 예시입니다. 파일명 충돌 시 '(2)' suffix가 추가될 수 있습니다.

```bash
# Bash mv inbox/{file}.md {destination}/{sub?}/{file}.md
```

--------------------------------

### Step 4 B모드: 서사 변경 시 상호작용

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/classify/references/workflow.md

Step 4에서 서사 변경이 필요할 경우, 사용자에게 확인을 요청하는 B모드 메시지 형식입니다. 개요, summary, aliases 수정 필요성을 알립니다.

```markdown
## 관련 페이지 수정 제안 (✋)

- 대상: [[{target_page}]]
- 변경점:
  - `## 개요` 섹션 업데이트
  - `summary` 필드 갱신
  - `aliases` 추가

수정해도 될까요?
```

--------------------------------

### Step 2 출력: 최종 폴더 및 분류 결정

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/classify/references/workflow.md

각 inbox 항목에 대한 최종 목적지, 하위 폴더, 분류 신뢰도, 결정 근거를 나타내는 JSON 객체입니다. 'entities' 타입의 경우 'sub' 필드가 포함됩니다.

```json
{
  "destination": "sources",  # entities | sources | cases | references | notes
  "sub": null,               # entities일 때만 (people/brands/works/concepts/media)
  "confidence": "high",      # high | medium | low
  "basis": "외부 source URL + 인물 언급 → sources/"
}
```

--------------------------------

### Partial Answer Template

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/query/references/report-template.md

Use this template when related files are found but only partially resolve the query within hop 2 limits, or when token limits prevent full exploration. The 'Caution' section is crucial for indicating unresolved parts and reasons for limitations.

```markdown
## 부분 답변
{해소된 부분 요약}

## 근거

### [[{path1}]] (Phase 2: 본문 hop 2, L2)
> {발췌}

## 주의
- {해소되지 않은 부분 명시}
- 한도 도달: {hop 2 종료 | 토큰 5k 초과 | 폴백 _index.md까지 시도}
- 질의를 더 구체화하면 정답률이 오른다: {제안}

## 질의 해석
- 키워드: {extracted_keywords}
- Phase 1 매핑: {질문 유형} → {진입점}
- 사용 Phase: {0, 1, 2}
- 누적 토큰: {N}/5000

## log
- log.md에 query 이벤트 기록 완료 (참조 {N}건, 결과: 부분)
```

--------------------------------

### Final Completion Output Message

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/classify/references/workflow.md

Example of the final output message upon successful completion of the classify workflow, summarizing processed items, updates, and pending actions.

```text
✓ classify 완료

- 처리: {N}개
- 이동: sources({n1}), cases({n2}), references({n3}), notes({n4}), entities({n5})
- 신규 엔티티: {K}개 ({names})
- related 자동: {L}건
- 서사 변경 ✋: {S}건
- 갱신된 페이지: [[index]], [[log]]
- 보류: {M}개 (low confidence — 다음 호출 시 재시도)
```

--------------------------------

### Step 1 출력: Inbox 파일 정보

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/classify/references/workflow.md

Step 1의 스캔 결과로, 각 inbox 파일의 경로, type, 요약, 출처, 미디어 타입, 태그 정보를 포함하는 JSON 배열입니다.

```json
[
  {
    "path": "inbox/2026-04-23-newjeans-comeback.md",
    "type": "inbox",
    "summary": "NewJeans 컴백 인터뷰...",
    "source": "https://...",
    "media": "article",
    "tags": ["...", "..."]
  },
  ...
]
```

--------------------------------

### JS 렌더링 필요 페이지 경고

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/ingest/references/sources.md

일반 URL Fetch 시 JavaScript 렌더링이 필요한 페이지에 대한 경고 메시지입니다. Chrome MCP 경로로 재시도할지 묻습니다.

```text
⚠️ 이 페이지는 JS 렌더링이 필요해 보입니다.
Chrome MCP 경로로 재시도할까요?
```

--------------------------------

### Step 2 Return Value Format

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/ingest/references/media.md

This JSON structure represents the output of the media inference process, including a summary, points, the inferred media type, confidence level, and the basis for the inference. This format is used by Step 4 for frontmatter injection.

```json
{
  summary: "{80~120자 한 줄}",
  points: ["...", "..."],
  media: {media},
  media_confidence: high | medium | low,
  media_basis: "host:youtube.com" | "ext:.pdf" | "fallback:article"
}
```

--------------------------------

### Unknown Answer Template

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/query/references/report-template.md

Use this template when no file-back evidence is found after completing retrieval phases 0-2. It explicitly states that no evidence was found and provides suggestions for query reformulation or data ingestion.

```markdown
## 모름

볼트에서 "{question}"에 대한 근거를 찾지 못했다.

- Phase 0 (hot.md): 매칭 0
- Phase 1 (폴더 직행 — {진입점}): 매칭 0
- Phase 2 (hop 2 + 폴백): 매칭 0

### 질의 재구성 제안
1. {구체적 제안 1 — 예: 동의어·다른 표기 (`aliases:` 필드)}
2. {구체적 제안 2 — 예: 폴더 한정 변경 또는 제거}
3. {구체적 제안 3 — 예: 엔티티명을 더 넓은 범주로}

### 가능성
- 볼트에 아직 수집되지 않음 — `/ingest`로 소스 추가 후 재질의
- 엔티티는 있지만 inbox에 머물러 있음 — `/classify`로 분류 후 재질의
- 표기 불일치 — 한글·영문·별칭(`aliases:`) 중 다른 표기로 시도

## log
- log.md에 query 이벤트 기록 완료 (참조: 없음, 결과: 모름)
```

--------------------------------

### Type 마커 불일치 알림 메시지

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/classify/references/workflow.md

inbox 파일의 frontmatter 'type:' 필드가 'inbox'가 아닐 경우 발생하는 경고 메시지입니다. 파일 처리 방안에 대한 옵션을 제공합니다.

```markdown
⚠️ inbox/{file}.md: type 마커 불일치
- 현재: {type}
- 기대: inbox
- 원인 추정: 옛 ingest 결과 또는 수동 작성

이 파일을 어떻게 처리할까요?
1. type을 `inbox`로 강제 변경 후 분류 진행
2. 이 파일 건너뛰고 나머지 진행
3. 전체 중단 (수동 검토)
```

--------------------------------

### New Skill Definition (SKILL.md)

Source: https://context7.com/choi-sanadesu/memography/llms.txt

This is the minimum structure required for a new skill definition in SKILL.md. It includes the skill's name, a description, and the main skill body.

```yaml
---
name: review
description: 볼트 페이지 간 모순·상충 진술을 탐지하고 검토 리포트를 출력한다. "리뷰", "모순 탐지", "/review" 같은 표현이 나올 때 트리거.
---

# memography: review

스킬 본문...
```

--------------------------------

### R2 Generation Preview Presentation

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/ingest/references/batch.md

This markdown table previews the planned new pages to be generated, including their intended path, title, type, media, tags, and how the original source will be handled. It's used in the R2 review stage for user confirmation before execution.

```markdown
## R2 생성 예정 ({N}개 신규 페이지)

| # | 경로 | 제목 | type | media | 태그 | 원본 처리 |
|---|---|---|---|---|---|---|
| 1 | inbox/2026-04-23-newjeans-comeback.md | NewJeans 컴백 인터뷰 | inbox | article | [...] | URL (이동 없음) |
| 2 | inbox/2026-04-23-pdf-report.md | 2026 K-POP IR 리포트 | inbox | paper | [...] | inbox 옆 복사 |
| 3 | inbox/2026-04-23-moodboard.md | 스프링 무드보드 | inbox | image | [...] | raw/images/로 복사 |
| 4 | ... | ... | ... | ... | ... | ... |

---

**{N}개 전부 생성?** 제외: "#3 제외". 슬러그 수정: "#1 슬러그 'XX'로".
```

--------------------------------

### Batch Ingest Progress Update

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/ingest/references/batch.md

This format shows the real-time progress of the batch ingest execution phase, indicating which step is being processed for each source and its corresponding destination path. It uses checkmarks for completed steps and a gear icon for the current step.

```text
✓ [1/5] {title1} — inbox/2026-04-23-newjeans-comeback.md
✓ [2/5] {title2} — inbox/2026-04-23-pdf-report.md
⚙ [3/5] 처리 중...
```

--------------------------------

### Ingest Step 2: Summary and Media Inference Confirmation

Source: https://context7.com/choi-sanadesu/memography/llms.txt

After providing a source, Memography presents a summary and inferred media type for user confirmation. This step allows review and approval before proceeding with inbox page creation. The confidence level for media inference is also indicated.

```text
# Step 2: Summary + Media Inference Results (User Confirmation Requested)
# ---
# ## Source Summary
#
# - summary: Pitchfork review of Smerz's second full-length album, Believer. Analyzes the Norwegian electronic duo's experimental production and emotional narrative.
# - Key Points:
#   1. Use of space and silence throughout the album
#   2. Henriette Motzfeldt's vocal layering techniques
#   3. Connection to the Scandinavian electronic scene
# - Nature: Review
#
# **media:** article (confidence: high)
#
# Proceed with this direction?

```

--------------------------------

### Chrome MCP 미연결 환경 메시지

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/ingest/references/sources.md

Chrome MCP가 연결되지 않은 환경에서 URL을 처리할 때 표시되는 경고 메시지입니다. 대안으로 코워크 환경 재실행 또는 로컬 파일로 저장 후 재호출을 제안합니다.

```text
⚠️ Chrome MCP 연결 없음

이 URL({host})은 브라우저 세션이 필요합니다.

대안:
1. 코워크·Chrome 연결된 환경에서 재실행
2. 브라우저에서 직접 열고 본문 복사
   → 로컬 파일로 저장 (예: ~/Downloads/{이름}.md)
   → `/ingest {파일경로}`로 재호출
```

--------------------------------

### Ingest Step 5: Completion Output

Source: https://context7.com/choi-sanadesu/memography/llms.txt

Upon successful completion of the ingest process, Memography provides a confirmation message. It lists the newly created page, its determined type and media, notes that the original source was only referenced, and indicates that the `log.md` file has been updated. It also suggests the next step of batch processing inbox items with `/classify`.

```text
# Step 5: Completion Output
# ✓ Ingest complete
#
# - New Page: [[inbox/2025-04-28-smerz-believer-pitchfork]]
# - type: inbox (to be confirmed by classify)
# - media: article (high)
# - Original Source Handling: URL reference only
# - Updated Page: [[log]]
# - Next Step: Can batch process inbox with `/classify`

```

--------------------------------

### Index Update Format for Moved Items

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/classify/references/workflow.md

Markdown format for adding an entry to index.md for each moved item, including its path, title, and a short summary.

```markdown
- [[{path}|{title}]] — {summary 짧게}
```

--------------------------------

### Step 3 출력: 파일 처리 결과

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/classify/references/workflow.md

각 inbox 항목이 성공적으로 처리되었음을 나타내는 메시지 형식입니다. 원본 경로, 새 경로, type 변경 정보를 포함합니다.

```text
✓ {original_path} → {new_path} (type: inbox → {new_type})
```

--------------------------------

### Batch Mode User Prompt for Media Confirmation

Source: https://github.com/choi-sanadesu/memography/blob/main/skills/ingest/references/workflow.md

This prompt is presented to the user in batch mode to confirm the inferred media type and its confidence level. It allows the user to provide feedback or override the system's classification.

```markdown
## 소스 요약

- summary: {80~120자}
- 핵심 논점:
  1. {point}
  2. {point}
  ...
- 성격: {기사 | SNS | 영상 | ...}

**media:** {media} (confidence: {level})
{media confidence low/medium일 때만 추가:}
- 근거: {host/확장자 — 예: "host:youtube.com", "ext:.pdf"}
- 대안: {alt_media}

이 방향으로 진행해도 될까요? 수정할 부분 있으면 지시해 주세요. (분류 보류 원하면 "media: null"로 남길 수 있습니다)
```

--------------------------------

### Claude Plugin Manifest (plugin.json)

Source: https://context7.com/choi-sanadesu/memography/llms.txt

The plugin.json file defines the metadata for the Claude plugin, including its name, version, description, and author.

```json
{
  "name": "memography",
  "version": "0.1.0",
  "description": "Memography Obsidian 볼트 운영 플러그인. inbox 기반 ingest·classify 2단계 워크플로우 + lint·query 운영 스킬.",
  "author": {
    "name": "choi-sanadesu"
  }
}
```

--------------------------------

### Ingest URL or File Path into Inbox

Source: https://context7.com/choi-sanadesu/memography/llms.txt

Use the `/memography:ingest` command to collect new sources into your vault's inbox. It supports URLs (HTTP/HTTPS) and file paths, creating a new page in `inbox/{YYYY-MM-DD}-{slug}.md` and logging the event. The process includes user confirmation before creating new pages, adhering to the B-mode principle.

```bash
# Ingest a single URL
/memography:ingest https://pitchfork.com/reviews/albums/smerz-believer

# Batch ingest (multiple sources simultaneously)
/memography:ingest https://example.com/article1 https://youtube.com/watch?v=xxx ~/Downloads/ref.pdf
```

--------------------------------

### Query Vault with Natural Language

Source: https://context7.com/choi-sanadesu/memography/llms.txt

This read-only skill answers natural language queries by referencing actual files in the vault. It follows a specific retrieval protocol, searching through phases from hot.md to log.md. If no evidence is found, it explicitly states 'Unknown' rather than guessing.

```bash
# Single natural language query
/memography:query What is Smerz?

# Query specifying a folder
/memography:query Artists from Copenhagen in the entities/people/ folder

# Query specifying a wiki link
/memography:query List of source pages related to [[entities/people/NewJeans]]
```

```markdown
# --- Answer Example (Confirmed) ---
# **Smerz** (`entities/people/Smerz.md`)
# 
# > "An electronic duo from Oslo, Norway, consisting of Henriette Motzfeldt and Catharina Stoltenberg."
# — entities/people/Smerz.md, frontmatter summary
# 
# Related pages:
# - [[sources/2025-04-28-smerz-believer-pitchfork]] — Pitchfork Believer review
# - [[entities/works/Believer]] — 2024 full-length album 2
# 
# log.md record completed

# --- Answer Example (Unknown) ---
# **Unknown**
# 
# Could not find file-backed evidence after searching through Phase 0-2.
# 
# Suggested query rephrasing:
# 1. Specify directly with a wiki link like `[[entities/people/Smerz]]`
# 2. Retry with broader keywords like "Scandinavian artist"
# 3. Ingest relevant sources first using `/memography:ingest {url}`
# 
# log.md record completed
```