Skip to main content
command CX Codex

obsidian-checkin — Claude Code command for checking notes

The obsidian-checkin command asks the vault owner to confirm aging facts in notes and records their replies. Use it to review details about people or projects that may have changed.

MIT ★ 5,412 updated 2026-10-24

Install into Claude Code · Plugin marketplace All methods and platforms ↓
/plugin marketplace add eugeniughelbur/obsidian-second-brain
/plugin install obsidian-second-brain@obsidian-second-brain
Open on GitHub

What obsidian-checkin does

The command selects notes with older facts and asks the owner to review them. It opens up to three cards at a time; replies confirming a fact, skipping it, or reporting news follow different paths. When details change, the note is edited before the card is marked complete.

Open questions live in .checkin.json, so a dropped message can be sent again. The ask, nudge, answer, and status modes cover starting a check-in, reminding the owner, processing replies, and listing open cards. It works with an Obsidian vault path and reads _CLAUDE.md first when that file exists.

Messages about health, family, or money go to the owner directly, never to a group chat. The command runs only when requested or on the owner’s schedule, not because a conversation happens to mention an old note.

What it helps with

  • find notes with facts older than a chosen age
  • ask the owner whether details about a person or project still hold
  • record a reply in the note and mark when it was checked
  • list open questions and send their reminders again

Example prompts for the agent

  • Check the old notes in my Obsidian vault.
  • Show me the open check-in cards.
  • The first item is unchanged.

How to install obsidian-checkin

1Plugin marketplace

/plugin marketplace add eugeniughelbur/obsidian-second-brain
/plugin install obsidian-second-brain@obsidian-second-brain

2Manually: put the files in place

.claude/commands/obsidian-checkin.md

Call it with /obsidian-checkin. For all projects use ~/.claude/commands/

Questions about obsidian-checkin

How do I install obsidian-checkin in Claude Code?

Run /plugin marketplace add eugeniughelbur/obsidian-second-brain, then run /plugin install obsidian-second-brain@obsidian-second-brain. The command file path is .claude/commands/obsidian-checkin.md.

What does obsidian-checkin do?

It asks the vault owner to review aging facts in notes, records their answers, and reports questions that remain open.

Where does obsidian-checkin send its questions?

Only directly to the owner. It must not send check-in messages or replies to group chats, team channels, or logs that others can read.

Source text

obsidian-checkin.md · MIT
---
description: Ask the owner whether old people and project notes still hold - shows what each note says and when it was last checked, re-asks until answered, saves the news
category: vault
trigger-mode: explicit
triggers_en: ["vault check-in", "check my old notes", "which notes are out of date", "ask me about stale notes"]
triggers_de: ["Vault-Check-in", "prüf meine alten Notizen", "welche Notizen sind veraltet", "frag mich nach veralteten Notizen"]
triggers_es: ["revisión de la bóveda", "revisa mis notas viejas", "qué notas están desactualizadas", "pregúntame por notas viejas"]
triggers_pt: ["check-in do cofre", "revise minhas notas antigas", "quais notas estão desatualizadas", "me pergunte sobre notas antigas"]
triggers_zh: ["知识库核对", "检查我的旧笔记", "哪些笔记过时了", "问我旧笔记有没有更新"]
---

Use the obsidian-second-brain skill. Execute `/obsidian-checkin $ARGUMENTS`:

A fact check compares notes with other notes, so news that never reached the vault (a bill settled, a move, a new job) leaves a note stale and no lint can see it. The owner is the only source. This command asks them, a few notes at a time, and saves what they answer.

The argument picks the mode: `ask` (default), `nudge`, `answer <the owner's reply>`, or `status`. The deterministic work lives in `scripts/checkin.py` (run it from the skill root: its absolute path was given at session start as **Skill root**; substitute it for `SKILL_ROOT`). Every call takes `--path <vault>`. The open questions live in `<vault>/.checkin.json`, never in the chat, so a question the host drops is sent again on the next run until it is answered.

1. Read `_CLAUDE.md` first if it exists in the vault root, for the vault path and privacy rules.

## Mode: ask (once a day)

2. Run `uv run --directory "SKILL_ROOT" scripts/checkin.py --path <vault> pick`. It returns up to 3 notes whose facts are older than 14 days (`--days N` to change), never more than 3 open at once (`--max-open N`). People tagged `family`, `relationship: strong` and `checkin: priority` come first, then active projects, then the most linked notes. Notes with a closed `status`, `checkin: never`, an open card, or a skip in the last 14 days are left out. An empty list means nothing is due: send nothing and stop.
3. For each picked note, read the note itself (the `excerpt` is a hint, not the source) and write **1 to 3 facts**:
   - One short line each, under 140 characters, in plain words.
   - Taken from the note only. Never add, guess, or round a fact. Keep dates and amounts exactly as written.
   - Prefer facts that can change: job, where they live, status, money, health, who they are with. Skip timeless ones like a birth date.
4. Open each card: `checkin.py --path <vault> ask --note "<path>" --fact "..." --fact "..."`.
5. Run `checkin.py --path <vault> message` and send its output to the owner **verbatim**, as one message. Do not add commentary, a summary, or a second message. A host may put its own marker (for example a notification emoji) in front.

## Mode: nudge (every few hours while the owner is awake)

6. Run `checkin.py --path <vault> message`. If it prints nothing, nothing is open: send nothing. Otherwise send the output verbatim. It repeats each card in full (name, date of the facts, the facts) with a reminder count, because the owner will not remember what an earlier message said. Never send a bare "still waiting" without the cards.

## Mode: answer (whenever the owner replies, even days later)

7. Run `checkin.py --path <vault> status` to see the open cards. Numbers in the reply refer to the last message sent (`last_message` in `.checkin.json`), so `--number N` resolves them.
8. Read the reply. It may cover several cards: `1 same`, `2 moved to Rotterdam`, `skip 3`, one per line or comma-separated. Free text with no number answers the only open card, when exactly one is open. If it is ambiguous, ask once which number it is for.
9. For each answer:
   - **same**: `checkin.py --path <vault> done --number N --outcome same`. This stamps `last-checked` with today, and the note stays quiet for 14 days.
   - **skip**: `done --number N --outcome skipped`. Nothing is stamped. The note can come back after 14 days.
   - **news**: update the note first, then `done --number N --outcome updated`.
     - Rewrite the stale claim where it lives. Do not only append. When the old value is still useful as history (an original bill before a settlement), keep it and label it as the earlier value.
     - Stamp the new claim: `(stated by <owner> YYYY-MM-DD)`, confidence `stated`. Anything the reply leaves out (a name, a date, an amount) is `TBD`, never guessed.
     - Set `updated:` to today on every note whose facts changed. Do NOT touch `updated:` on a note you only linked or formatted, or it hides from the check-in for a cycle.
     - Propagate: every person, project, or place in the news gets the change on its own note too, as `/obsidian-save` would, with `[[wikilinks]]` both ways.
10. Check every note you wrote: `uv run --directory "SKILL_ROOT" scripts/check_notes.py --path <vault> --fix "<path>" ...`. It runs the same AI-first check Claude Code runs on every write, fixes banned characters, and exits 1 on anything left. Fix what it reports before you finish.
11. Reply with one short line, for example `Saved: 1 same, 2 updated (Ana, moved to Rotterdam).` If cards are still open, say nothing more: the next nudge will carry them.

## Mode: status

12. Run `checkin.py --path <vault> status` and show the open cards as a short list: name, date of the facts, times sent.

## Rules

- **Private channel only.** Check-in messages and answers can hold family, health, and money facts. Send them only to the owner directly, never to a group chat, a team channel, a post, or a log that others read.
- **Short.** One message per run, at most 3 cards in it. The script does the formatting so every host sends the same shape.
- **Explicit only.** Never start a check-in because a conversation drifted near an old note. It runs when the owner asks or on the owner's own schedule.
- **Per note control.** `checkin: never` in a note's frontmatter keeps it out for good. `checkin: priority` moves it to the front.
- After each mode, append to the operation log: if `Logs/` exists write `**HH:MM** - checkin <mode> | N asked, N answered, N open` to `Logs/YYYY-MM-DD.md`; otherwise append `## [YYYY-MM-DD] checkin <mode> | N asked, N answered, N open` to `log.md`. Names and facts stay out of the log line.

## Scheduling

Hosts with a scheduler run `ask` once a day and `nudge` every 2 to 4 hours inside the owner's waking hours. Hosts without hooks (Grok Bot, cron agents) must run step 10 after every vault write, not only here: it is their write-time check.

---

**AI-first rule:** Every note created or updated by this command MUST follow `references/ai-first-rules.md` - `## For future agent` preamble, rich frontmatter (`type`, `date`, `tags`, `ai-first: true`, plus type-specific fields), recency markers per external claim, mandatory `[[wikilinks]]` for every person/project/concept referenced, sources preserved verbatim with URLs inline, and confidence levels where applicable. If that path does not resolve from your working directory, search upward for it; if you still cannot read it, say so before writing instead of producing a note that silently skips the rule. The vault is for future agent retrieval - not human reading.

**Anti-fabrication:** Search exhaustively before claiming any note, person, or file is absent - false absence is the most common failure mode - and never invent facts, entities, or dates (mark unknowns as `TBD`). See the anti-fabrication and search-completeness hard rules in `references/ai-first-rules.md`.

© Eugeniu Ghelbur · MIT · eugeniughelbur/obsidian-second-brain

Similar slash commands

All slash commands for Notes, tasks and projects →

From the same set51

The whole set →
obsidian-reconcile obsidian-reconcile is a Claude Code command that finds contradictions in an Obsidian vault and resolves them automatically or flags them for review. It scans the full vault or a specific topic. CX Codex +1 ★ 5,412repo stars obsidian-visualize The obsidian-visualize command maps your Obsidian vault visually: shows how notes connect, identifies hubs, and highlights isolated entries. Use it to see your second brain's shape at a glance. CX Codex +1 ★ 5,412repo stars obsidian-distill obsidian-distill condenses a long note or article into key claims, each tagged with the exact source block it came from — paragraph, section, or timestamp. CX Codex +1 ★ 5,412repo stars obsidian-merge obsidian-merge merges near-duplicate Obsidian notes without deleting files. It preserves links, unions frontmatter, and converts the retired note into a redirect stub. CX Codex +1 ★ 5,412repo stars research-deep research-deep scans your Obsidian vault to map what you already know, identifies missing or stale information, fills gaps via Perplexity or Grok, synthesizes the findings, and updates your notes on people, projects, and ideas automatically. CX Codex +1 ★ 5,412repo stars create-command create-command scaffolds a new slash command for Obsidian through an interactive interview, removing the need to write Markdown or frontmatter by hand. CX Codex +1 ★ 5,412repo stars obsidian-init obsidian-init scans an Obsidian vault and creates _CLAUDE.md, index.md, and a log pointer. Use it to prepare vault guidance and a note catalog for Claude. CX Codex +1 ★ 5,412repo stars obsidian-health obsidian-health scans your Obsidian vault for contradictions, concept gaps, stale claims, and structural issues. Reports findings grouped by severity and spins up specialized agents to triage each category. CX Codex +1 ★ 5,412repo stars

Project sponsors

Companies that keep this analytics open