wiki-digest — скилл для Claude Code: дайджест знаний из Obsidian
Собирает читаемый отчёт о том, что вы узнали и обновили в вики за день, неделю или месяц. Пригодится для регулярного разбора заметок.
npx skills add Ar9av/obsidian-wiki --skill wiki-digest -a claude-codeЧто делает wiki-digest
Скилл читает log.md и hot.md в вашем Obsidian-хранилище, отбирает страницы, созданные или изменённые за выбранный период, и превращает их в структурированный markdown-отчёт. Внутри — заголовки с выводами, таблицы новых и обновлённых страниц, частотные теги, связи между категориями и список материалов, к которым стоит вернуться.
Период задаётся словами: «daily», «weekly», «monthly», ISO-дата или число дней. По умолчанию берётся последняя неделя. Если активных страниц меньше пяти, скилл останавливается и предлагает расширить окно — это повод проверить, пополняете ли вы вики вообще.
Для работы нужен путь к хранилищу: скилл разрешает конфиг по протоколу из llm-wiki/SKILL.md — сначала инлайн-переопределение @name, потом поиск .env вверх по дереву каталогов, затем глобальный конфиг. Без OBSIDIAN_VAULT_PATH он попросит настроить его. Ещё читается _meta/taxonomy.md, чтобы отметить теги, которых в таксономии ещё нет.
Какие задачи решает
- собрать недельный дайджест по свежим заметкам
- найти темы, которые чаще всего всплывали за месяц
- отметить новые теги, не попавшие в таксономию
- выбрать старые страницы, которые стоит перечитать
- посчитать черновики и необработанные заметки в
_raw/
Примеры запросов агенту
- сделай weekly дайджест по моей вики
- покажи, что нового появилось за последние 14 дней
- собери monthly digest и отметь новые теги
Как установить wiki-digest
1Через Skills CLI (Vercel)
npx skills add Ar9av/obsidian-wiki --skill wiki-digest -a claude-code2Вручную: положить файлы
~/.claude/skills/wiki-digest/SKILL.mdПапка скилла целиком; для одного проекта — .claude/skills/wiki-digest/
1Через Skills CLI (Vercel)
npx skills add Ar9av/obsidian-wiki --skill wiki-digest -a codex2Вручную: положить файлы
.agents/skills/wiki-digest/SKILL.mdДля всех проектов — ~/.agents/skills/
1Через Skills CLI (Vercel)
npx skills add Ar9av/obsidian-wiki --skill wiki-digest -a cursor2Вручную: положить файлы
.cursor/skills/wiki-digest/SKILL.mdCursor читает и .agents/skills/, и .claude/skills/
1GitHub CLI
gh skill install Ar9av/obsidian-wiki wiki-digest2Вручную: положить файлы
.github/skills/wiki-digest/SKILL.md1Через Skills CLI (Vercel)
npx skills add Ar9av/obsidian-wiki --skill wiki-digest -a gemini-cli2Командой платформы
gemini skills install https://github.com/Ar9av/obsidian-wiki --path .skills/wiki-digestВручную: положить файлы
.devin/skills/wiki-digest/SKILL.mdСтарый путь .windsurf/skills/ тоже читается
Через Skills CLI (Vercel)
npx skills add Ar9av/obsidian-wiki --skill wiki-digest -a opencodeClaude.ai
Упакуйте папку в zip и загрузите в Settings → Capabilities → Skills
По шагам, с командами для каждой платформы: Как установить скилл в Claude Code: пять способов
Вопросы о wiki-digest
Как установить wiki-digest в Claude?
Выполните npx skills add Ar9av/obsidian-wiki --skill wiki-digest -a claude-code. Скилл встанет в ~/.claude/skills/wiki-digest/SKILL.md.
Что делает wiki-digest?
Формирует дайджест знаний за день, неделю или месяц: новые и обновлённые страницы, частые темы, интересные связи и рекомендации по перечитыванию.
Что нужно для работы скилла?
Путь к Obsidian-хранилищу в OBSIDIAN_VAULT_PATH и формат ссылок в OBSIDIAN_LINK_FORMAT. Конфиг разрешается по протоколу из llm-wiki/SKILL.md.
Исходный текст
SKILL.md · MIT
---
name: wiki-digest
description: >
Produce a day/week/month knowledge digest summarizing what was learned, updated, and connected in the wiki. Use for readable recent-learning summaries; wiki-status reports source and ingestion state instead.
---
# Wiki Digest — Knowledge Newsletter Generator
You are generating a human-readable digest of recent wiki activity: what was learned, what was updated, what themes are emerging, and what's worth reviewing. This skill summarizes *knowledge*, not sources — think of it as a weekly review session, not an ingestion status report.
## Before You Start
**Writing profile:** Before drafting or rewriting natural-language Markdown, read and apply the `Writing Profile Resolution` section in `llm-wiki/SKILL.md`. Framework schema, provenance, safety, and operation-specific requirements take precedence.
`WRITING.md` preferences apply only to newly drafted or rewritten natural-language Markdown; preserve source content and structured records.
1. **Resolve config** — follow the Config Resolution Protocol in `llm-wiki/SKILL.md` (inline `@name` override → walk up CWD for `.env` → global config → prompt setup). This gives `OBSIDIAN_VAULT_PATH` and `OBSIDIAN_LINK_FORMAT`.
2. **Parse the period** from the user's request:
- "daily" / "today" / "yesterday" → last 24 hours
- "weekly" / "this week" / no argument (default) → last 7 days
- "monthly" / "this month" → last 30 days
- ISO date like "since 2026-05-01" → pages updated since that date
- Explicit number like "last 14 days" → that many days
3. Read `$OBSIDIAN_VAULT_PATH/log.md` — last 200 lines — for entries within the period (timestamps are ISO-8601 prefixed lines).
4. Read `$OBSIDIAN_VAULT_PATH/hot.md` for current session context.
5. If `$OBSIDIAN_VAULT_PATH/_insights.md` exists, read its **Anchor Pages** table — you'll use it later to identify which new pages became hubs.
## Step 1: Collect Pages Active in the Period
Glob all `.md` files under `$OBSIDIAN_VAULT_PATH`. Skip special/system files:
- `index.md`, `log.md`, `hot.md`, `AGENTS.md`, `_insights.md`
- Anything under `_meta/`, `_archives/`, `_raw/`
- Journal digest pages themselves (`journal/digest-*.md`)
For each remaining page, read its frontmatter:
- `created` — when the page was first written
- `updated` — when it was last modified
Classify:
- **New pages**: `created` is within the period
- **Updated pages**: `updated` is within the period but `created` is before it
- **Unchanged**: neither date falls in the period → skip
If fewer than 5 pages were active, note it and offer to widen: *"Only 3 pages were active in the last 7 days — want a monthly digest instead?"* Stop here unless the user says to continue.
For each active page, collect: `title`, `category`, `tags`, `summary` (frontmatter field), `lifecycle`, any `^[ambiguous]` or `^[inferred]` markers in the body.
## Step 2: Identify Themes
From all active pages' tags, tally theme frequency:
```
For each tag across new + updated pages:
count how many active pages carry it
Sort descending, take top 5
```
Also read `$OBSIDIAN_VAULT_PATH/_meta/taxonomy.md` (if it exists). Flag any tag from step 1 that **does not appear** in the taxonomy — these are new vocabulary words that emerged this period.
Note which categories grew most (concepts/, entities/, skills/, synthesis/, references/, etc.).
## Step 3: Find Notable New Connections
Scan new and updated pages for cross-category wikilinks — links that bridge different knowledge layers. These are the most intellectually interesting outputs of the period.
For each active page, extract all `[[wikilink]]` targets. Classify each link by the target's category prefix. Flag links that cross categories (e.g., a `concepts/` page linking to an `entities/` page, or a `synthesis/` page bridging two topics).
Rank candidates by interestingness:
- **+3** if the link is across two categories that rarely connect (use `_insights.md` bridge data if available)
- **+2** if the target page is a top-10 hub (per `_insights.md` anchors)
- **+2** if the link appears in a `synthesis/` page (deliberate cross-cutting)
- **+1** if the source page is marked `^[inferred]` (synthesized connection, not directly stated)
Take the top 3–5 connections. Write each as a plain-English sentence: not just "A → B" but *why* the connection is interesting.
## Step 4: Surface Open Threads
Scan active pages and `_raw/` for unresolved work:
- **Drafts**: pages with `lifecycle: draft` or `lifecycle: stub`
- **Ambiguous claims**: count `^[ambiguous]` markers across all active pages (don't list every one — just the count and which pages have the most)
- **Unstaged notes**: count files in `$OBSIDIAN_VAULT_PATH/_raw/` (anything here hasn't been promoted)
- **Taxonomy gaps**: tags from Step 2 that aren't in `_meta/taxonomy.md`
## Step 5: Choose Recommended Re-reads
From the *existing* (pre-period) pages, identify 2–3 worth revisiting given this week's new context.
Heuristic: find pre-period pages that share the most tags with the active pages from Step 1. These are foundational pages whose topic was extended this period — the new pages build on them but the user may not have revisited the foundation.
Also include any pre-period page that now has 2+ new incoming links from active pages (it just became more connected — a sign it's load-bearing).
Write each recommendation with a concrete reason: *"[[concepts/attention-mechanism]] — your foundational page; three new papers ingested this week all extend it"*, not just the page title.
## Step 6: Generate the Digest
Produce a structured, scannable markdown report. The Headlines section is the most important — it should feel like the opening of a good newsletter, synthesizing actual insight rather than listing page names.
Apply the link format from `llm-wiki/SKILL.md` (Link Format section) using `OBSIDIAN_LINK_FORMAT`. Default is `[[wikilink]]`.
```markdown
# Wiki Digest — [Period Label]
> [N new pages · M updated pages · period: YYYY-MM-DD to YYYY-MM-DD]
## Headlines
- [Concrete insight #1 — synthesize the actual knowledge, not just "learned about X"]
- [Concrete insight #2]
- [Concrete insight #3]
## New Knowledge
### New pages ([count])
| Page | Category | Summary |
|---|---|---|
| [[concepts/foo]] | concept | One-sentence summary from frontmatter |
| [[entities/bar]] | entity | One-sentence summary |
### Notable updates ([count])
| Page | What changed |
|---|---|
| [[skills/react-hooks]] | Added patterns for useCallback with async effects |
*(If no updates, omit this subsection.)*
## Emerging Themes
- **#[tag]** ([N pages]) — [One sentence on why this topic was active]
- **#[tag]** ([N pages]) — [...]
- **#[NEW TAG]** ([N pages]) ⭐ *New vocabulary — not yet in taxonomy*
Most active category: **[category/]** ([N pages added or updated])
## Key Connections Made
- [[concepts/A]] → [[entities/B]] — [Plain-English reason this connection is interesting]
- [[synthesis/X]] created — bridges [[concepts/Y]] and [[concepts/Z]] for the first time
- *(up to 5 connections)*
## Open Threads
- **Drafts to compile** ([count]): [[concepts/foo]], [[concepts/bar]] — still in draft lifecycle
- **Ambiguous claims**: [N] `^[ambiguous]` markers across [M] pages — run `/wiki-synthesize` to resolve
- **Unstaged notes**: [N] files in `_raw/` — run `/wiki-ingest _raw/` to promote them
- **Taxonomy gaps**: Tags `#newtag1`, `#newtag2` used but not in taxonomy — run `/tag-taxonomy`
*(Omit any subsection where count is 0.)*
## Recommended Re-reads
- [[concepts/X]] — [Specific reason: "3 new papers this week all extend this concept"]
- [[synthesis/Y]] — [Specific reason: "2 new pages created this week reference it"]
- [[skills/Z]] — [Specific reason: "now has 4 new incoming links — it's become a hub"]
---
*Generated by wiki-digest · [TIMESTAMP] · [N pages scanned in [VAULT_PATH]]*
```
**Visibility**: If a page is tagged `visibility/pii`, exclude it from all tables and connection lists (but count it in the totals, noted as "+ N private"). If the user explicitly says "include private pages" or "full digest", include them normally.
## Step 7: Output & Optionally Save
**Default (chat output):** Print the digest directly. At the end, ask:
*"Want me to save this as `journal/digest-YYYY-MM-DD.md`?"*
**If user prefixed with "save" or "write"** (e.g., `/wiki-digest save` or "generate and save my weekly digest"):
- Write to `$OBSIDIAN_VAULT_PATH/journal/digest-YYYY-MM-DD.md` (weekly/monthly) or `journal/digest-YYYY-MM-DD-daily.md` (daily)
- Add frontmatter:
```yaml
---
title: "Wiki Digest — [Period Label]"
category: journal
tags: [digest, meta/review]
sources: []
created: TIMESTAMP
updated: TIMESTAMP
summary: "Weekly knowledge digest: [N new, M updated pages]. Top themes: [tag1], [tag2]."
---
```
- Update `index.md` with the new entry under Journal
- Do **not** add to `.manifest.json` (digests aren't source ingestions)
Either way, append to `log.md`:
```
- [TIMESTAMP] DIGEST period="7d" new_pages=N updated_pages=M themes=T connections=C saved=false
```
## Edge Cases
| Situation | Handling |
|---|---|
| Fewer than 5 active pages | Offer to widen the period; proceed only if user confirms |
| Empty vault (no pages at all) | Tell the user to run an ingest first; stop |
| No `_meta/taxonomy.md` | Skip taxonomy gap check; omit that line from Open Threads |
| No `_insights.md` | Skip hub-based scoring in Step 3; still produce connections section |
| All pages are `visibility/pii` | Report "N private pages active this period" with no details; offer full mode |
| Period spans a wiki rebuild | Note it in the digest: "Wiki was rebuilt during this period — page dates reflect post-rebuild state" |
## Notes
- **Headlines are the payoff.** Don't list page titles — synthesize the actual learning. If someone learned about attention mechanisms this week, the headline should capture the insight, not just say "added 3 transformer pages".
- **Be concrete about re-reads.** "This page is relevant" is useless. "3 of this week's papers all cite the same claim in this page" is actionable.
- **This skill only reads.** The only writes are the optional journal page, and the `log.md` append. It does not modify existing wiki pages.
- **Don't duplicate wiki-status.** If the user asks "what needs ingesting" or "what's the delta", route to `wiki-status`. This skill answers "what did I learn", not "what's pending".
## QMD Refresh After Vault Writes
QMD is a search index, not the source of truth. If `$QMD_WIKI_COLLECTION` is empty or unset, skip this step. Run it only after this skill has written or rewritten vault markdown. If QMD refresh fails, do not roll back the vault changes; report the QMD status separately.
Use `$QMD_CLI` if set; otherwise use `qmd`.
```bash
${QMD_CLI:-qmd} update
```
If the output says vectors are needed or embeddings may be stale, run:
```bash
${QMD_CLI:-qmd} embed
```
Verify the collection with either:
```bash
${QMD_CLI:-qmd} ls "$QMD_WIKI_COLLECTION"
```
or, when a specific page path is known:
```bash
${QMD_CLI:-qmd} get "qmd://$QMD_WIKI_COLLECTION/<page>.md" -l 5
```
Record one of:
- `QMD refreshed: update + embed + verified`
- `QMD refreshed: update only + verified`
- `QMD skipped: QMD_WIKI_COLLECTION unset`
- `QMD skipped: qmd CLI unavailable`
- `QMD failed: <short error summary>`
© Ar9av · MIT · Ar9av/obsidian-wiki