ObsidianOS: Work

AutoGineer overview, from the brain’s own files

The system operates as a markdown-based vault where AI agents interact with files via a set of predefined skills located in the .agents/skills/ directory. These skills are triggered by slash commands that execute sequences of operations, such as fetching meeting transcripts from Google Docs, organizing files into specific folders, extracting action items, and committing changes to version control. The vault uses Obsidian Bases for data views and queries, and relies on external CLI tools for tasks like Gmail transcript processing and semantic search indexing via QMD.

The agent reads

• .cursor/mcp.json evidence: CLAUDE.md
• .cursor/rules/ evidence: README.md
• AGENTS.md evidence: README.md
• CLAUDE.md evidence: README.md
• OpenCode.md evidence: README.md
• Meetings.base evidence: README.md

The agent writes

• .gitattributes evidence: AGENTS.md
• .agents/skills/ evidence: CLAUDE.md
• USER.md evidence: README.md
• Meetings/ evidence: README.md
• Meetings/Engineering/ evidence: README.md

Agent instructions that run or fetch something (71)

• .agents/skills/cache-notes/SKILL.md: If the target file has an **empty `Notes:` property** (no URLs yet), run **`/mail-transcripts`** first when doing a full wrap (see [meeting](../meeting/SKILL.md) Mode C) — it may populate `Notes:` from the Gemini inbox mail.
• .agents/skills/cache-notes/SKILL.md: Run **`gws`** in the terminal to fetch the document (read-only).
• .agents/skills/cache-notes/SKILL.md: **Implementation**: Use the [otter-fetch](../otter-fetch/SKILL.md) skill — ask the user to paste the JSON from `GET .../speech?otid=...`, then run `parse_otter.js` (or parse inline) to produce the callouts above and append them.
• .agents/skills/cache-notes/SKILL.md: To check whether the body already has an `## 🤖 AI Notes` section without reading the whole file, run `obsidian outline path="<file>" format=json` and look for that heading — much cheaper than a full `Read` on large transcript files.
• .agents/skills/cache-notes/SKILL.md: **For each URL in `Notes:`**, fetch and parse by provider: · uses the network
• .agents/skills/cache-notes/SKILL.md: run `otter-fetch/parse_otter.js` or parse inline), then append the resulting callout block.
• .agents/skills/commit/SKILL.md: Run in parallel:
• .agents/skills/commit/SKILL.md: `git diff --staged --stat` — staged changes
• .agents/skills/commit/SKILL.md: `git diff --stat` — unstaged changes
• .agents/skills/commit/SKILL.md: `git status --short` — untracked files
• .agents/skills/commit/SKILL.md: `git log --oneline -5` — recent commits (for style and amend safety)
• .agents/skills/commit/SKILL.md: If unrelated files are staged, unstage them first (`git restore --staged <path>`).
• .agents/skills/commit/SKILL.md: Analyze the staged diff (`git diff --staged`) and write:
• .agents/skills/commit/SKILL.md: **Normal mode**: `git commit` with the confirmed message.
• .agents/skills/commit/SKILL.md: **Amend mode**: `git commit --amend` (see safety rules below).
• .agents/skills/commit/SKILL.md: **HEAD was created by you** in this conversation — check `git log -1`.
• .agents/skills/commit/SKILL.md: If `git status` fails, skip entirely without error.
• .agents/skills/defuddle/SKILL.md: If not installed: `npm install -g defuddle`
• .agents/skills/fill-participants/SKILL.md: Run [/note-status pending --step=participants](../note-status/SKILL.md#pending-mode) to discover notes missing `Participants:`, present them, and let the user select which to process.
• .agents/skills/fill-participants/SKILL.md: run `parse_otter.js` or parse inline).
• .agents/skills/fill-participants/SKILL.md: If `git status` fails, skip the commit step entirely without error.
• .agents/skills/followup-todos/SKILL.md: `/followup-todos` (no args) — Run [/note-status pending --step=todos](../note-status/SKILL.md#pending-mode) to show meetings missing follow-up extraction.
• .agents/skills/mail-transcripts/SKILL.md: **`/mail-transcripts`** — Run the Gmail → vault sync **once** (no arguments).
• .agents/skills/mail-transcripts/SKILL.md: From the **vault root**, run:
• .agents/skills/mail-transcripts/SKILL.md: python3 .scripts/gmail_ai_transcripts_to_meetings.py
• .agents/skills/mail-transcripts/SKILL.md: When `/mail-transcripts` is run **standalone**, offer [/commit](../commit/SKILL.md).
• .agents/skills/mail-transcripts/SKILL.md: When run **inside** `/meeting wrap` or `/meeting wrap pending`, the parent sequence commits once at the end (include any new/updated meeting files in that commit scope).
• .agents/skills/meeting/SKILL.md: **Run `/note-status pending [<dates>]`** — Discovers meeting files, filters to notes with any missing step, presents the status table, and prompts the user to select which meetings to wrap.
• .agents/skills/meeting/SKILL.md: **For each selected meeting**, run the Mode C wrap sequence starting at step 2 (`/cache-notes` → `/fill-participants` → `/followup-todos`).
• .agents/skills/migrate-tags/SKILL.md: `/migrate-tags --dry-run` — Show what would change without writing.
• .agents/skills/migrate-tags/SKILL.md: Read the file body, collect all `#tag` patterns (respecting word boundaries — `#tag` not `https://foo.com/path#section`). · uses the network
• .agents/skills/migrate-tags/SKILL.md: Run `obsidian property:read name=tags path="<file>"` on each migrated file to confirm the write.
• .agents/skills/note-status/SKILL.md: Run the standard workflow (Steps 1–4 above) with the given `[dates]` or `all` if omitted.
• .agents/skills/obsidian-bases/SKILL.md: | Function | Signature | Description | |----------|-----------|-------------| | `date()` | `date(string): date` | Parse string to date (`YYYY-MM-DD HH:mm:ss`) | | `now()` | `now(): date` | Current date and time | | `today()` | `today(): date` | Current date (time = 00:00:00) | | `if()` …[+244 chars] · uses the network
• .agents/skills/obsidian-cli/SKILL.md: Run `obsidian help` to see all available commands.
• .agents/skills/obsidian-cli/SKILL.md: Run JavaScript in the app context:
• .agents/skills/obsidian-cli/SKILL.md: Run `obsidian help` to see additional developer commands including CDP and debugger controls.
• .agents/skills/onboard-person/SKILL.md: Run `obsidian property:read name=FullName path="Teams/People/@<Name>.md"` to confirm the write.
• .agents/skills/organize-meetings/SKILL.md: The skill does not modify frontmatter — it only moves files (`git mv`) and updates the H1 inside when renaming.
• .agents/skills/organize-meetings/SKILL.md: **Otherwise, run qmd-vector_search** on the title — return top 5 most similar meeting paths.
• .agents/skills/organize-meetings/SKILL.md: `git mv "<old path>" "<new path>"` — preserves history.
• .agents/skills/otter-fetch/SKILL.md: It's a sub-skill: `/cache-notes` and `/fill-participants` invoke its workflow (and run `parse_otter.js`) when a meeting's `Notes:` contains an `https://otter.ai/u/...` URL.
• .agents/skills/otter-fetch/SKILL.md: **Ask the user** to copy the JSON response from `GET https://otter.ai/forward/api/v1/speech?otid={otid}` (e.g. from DevTools → Network while signed in, or from a signed-in request). · uses the network
• .agents/skills/otter-fetch/SKILL.md: | Use case | What to use | How | |----------|-------------|-----| | **cache-notes** | Callout markdown | Run `parse_otter.js` with JSON on stdin → stdout is `## 🤖 AI Notes` / `### Otter` block with `[!otter_notes]-`, `[!otter_todos]-`, `[!otter_transcript]-`.
• .agents/skills/otter-fetch/SKILL.md: node .agents/skills/otter-fetch/parse_otter.js < pasted.json
• .agents/skills/people-stubs/SKILL.md: Run `obsidian unresolved format=json` and filter to entries where the link target starts with `@`.
• .agents/skills/people-stubs/SKILL.md: Use `obsidian create path="Teams/People/@<Name>.md" content="<frontmatter>"` so Obsidian indexes it immediately (the next `/vault-health` run won't show the link as unresolved any more).
• .agents/skills/people-stubs/SKILL.md: - Or run /onboard-person <name> to pull metadata from external sources.
• .agents/skills/qmd/SKILL.md: !`qmd status 2>/dev/null || echo "Not installed: npm install -g @tobilu/qmd"`
• .agents/skills/qmd/SKILL.md: curl -X POST http://localhost:8181/query \ · uses the network
• .agents/skills/recap/SKILL.md: Run the following in parallel where possible:
• .agents/skills/recap/SKILL.md: For each selected `id`, run `gws gmail users messages get` with `"format":"full"`.
• .agents/skills/sprint-retro/SKILL.md: If you encounter one while gathering context, flag it and offer to `git mv` it into the canonical form (updating any `[[wikilinks]]` that reference the old name).
• .agents/skills/sprint-retro/SKILL.md: Confirm the resolved sprint number and `start..end` window with the user if anything is ambiguous (missing prior retro, irregular cadence, mid-sprint run).
• .agents/skills/sprint-retro/SKILL.md: **Naming consistency** — if you touch sibling planning/retro files with non-canonical names, normalize them via `git mv` and fix referencing wikilinks (see the convention section).
• .agents/skills/sync-upstream-obsidianos/SKILL.md: Shortcut to run [`.scripts/sync-upstream.sh`](../../../.scripts/sync-upstream.sh).
• .agents/skills/sync-upstream-obsidianos/SKILL.md: Verify `upstream` remote exists: `git remote get-url upstream`.
• .agents/skills/sync-upstream-obsidianos/SKILL.md: Verify working tree is clean (`git status --short`).
• .agents/skills/sync-upstream-obsidianos/SKILL.md: Run `.scripts/sync-upstream.sh --preview` and show the output (new commits and changed files).
• .agents/skills/sync-upstream-obsidianos/SKILL.md: On confirmation, run the full sync:
• .agents/skills/sync-upstream-obsidianos/SKILL.md: The final commit (`git log -1 --oneline`).
• .agents/skills/vault-health/SKILL.md: `/vault-health` — Run all checks, present a triage report.
• .agents/skills/vault-health/SKILL.md: Run only the requested subset;
• .agents/skills/workspace-mode/SKILL.md: No args → run `obsidian workspace format=json` and present the list with the current one starred.
• .agents/skills/workspace-mode/SKILL.md: `<name>` arg → run `obsidian workspace load name=<name>` (or use the Obsidian "Workspaces: Load workspace" command if `load` isn't a direct subcommand: `obsidian command name="Workspaces: Load workspace layout"`).
• .agents/skills/workspace-mode/SKILL.md: `save <name>` → run `obsidian command name="Workspaces: Save layout"` then prompt for the name in Obsidian's modal.
• AGENTS.md: **QMD** — Configured in [.cursor/mcp.json](.cursor/mcp.json) for vault search (MCP server `npx qmd mcp`).
• AGENTS.md: **Google Workspace** — Use the **`gws` CLI** ([Google Workspace CLI](https://github.com/googleworkspace/cli)) for Docs, Drive, Calendar, and Gmail in read-only workflows. · uses the network
• AGENTS.md: git remote add upstream <url-to-this-repo>
• AGENTS.md: ./.scripts/sync-upstream.sh # auto-configures merge driver on first run
• CLAUDE.md: Install and authenticate per [README.md](README.md) § Google Workspace CLI.

MCP servers (1)

• .cursor/mcp.json (qmd): npx qmd mcp