organize-meetings
Skill166 starsChanged 5 months agoMITView on GitHub
Triage Meetings/_inbox/ — move each file to the correct subfolder by finding adjacent/similar notes via qmd, applying learned routing conventions from memory, and confirming with the user.
In this repository29 other files it gives its agents.
- AGENTS.mdAGENTS.md
- .agents/skills/CLAUDE.mdCLAUDE.md
- CLAUDE.mdCLAUDE.md
- .cursor/rules/qmd-search.mdcCursor rule
- .cursor/rules/skill-conventions.mdcCursor rule
- .cursor/rules/skill-registry.mdcCursor rule
- cache-notesSkill
- commitSkill
- defuddleSkill
- fill-participantsSkill
- followup-todosSkill
- json-canvasSkill
- last-1on1Skill
- mail-transcriptsSkill
- meetingSkill
- migrate-tagsSkill
- note-statusSkill
- obsidian-basesSkill
- obsidian-cliSkill
- obsidian-markdownSkill
- onboard-personSkill
- otter-fetchSkill
- people-stubsSkill
- qmdSkill
- recapSkill
- sprint-retroSkill
- sync-upstream-obsidianosSkill
- vault-healthSkill
- workspace-modeSkill
The file
We split this skill into the rules it gives an agent
12 rules. About 0 are common, 4 are distinctive, and 8 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
8
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 41
If the candidate list is empty, print Nothing to organize.
- Very specific: line 45
Check memory/feedback_meeting-routing.md (if present).
- Very specific: line 61
All siblings end in - YYYY-MM-DD.md → keep date suffix.
- Very specific: line 62
All siblings use YYYY-MM-DD.md only (e.g.
- Distinctive: line 85
Ask the user: "all", "high-confidence only", "1,3", "none", or per-file corrections (e.g.
- Distinctive: line 92
If the new filename differs from the old, update the H1 inside the file.
- Very specific: line 96
When the user overrides a proposal ("don't put Pat/Me in One-on-ones, it goes to Tech Syncs"), this is feedback worth saving.
- Distinctive: line 115
Always include a Why: (the user's reasoning, if given) and How to apply: (the precondition + filename pattern).
- Very specific: line 121
When called standalone: commit with update: /organize-meetings — N files moved (or similar).
- Very specific: line 133
Always use git mv, never mv — history matters.
- Very specific: line 134
Never auto-move low confidence proposals, even in --auto mode.
- Distinctive: line 136
If qmd's index is stale (a recent batch of renames hasn't been reindexed), low-confidence proposals may be common — note this in the …
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Organize Meetings
Usage
/organize-meetings— SweepMeetings/_inbox/and propose destinations for every file./organize-meetings <path>— Categorize a single file (in or out of_inbox/)./organize-meetings --auto— Apply only high-confidence proposals (≥ 4/5 majority); leave the rest in_inbox/for next sweep.
When to run
| Caller | When |
| Standalone | Whenever you want to drain Meetings/_inbox/ (e.g. after /mail-transcripts created some stubs). |
/meeting wrap pending | Step 1.5 — between /mail-transcripts and pending discovery, so triaged files are included in the pending scan. |
Why this skill exists
The Gemini-emailed archiver (.scripts/gmail_ai_transcripts_to_meetings.py) only knows deterministic patterns: daily standups, sprint ceremonies, recurring team meetings, etc. Everything name-dependent — 1:1s with anyone, cross-team syncs, novel meeting titles — lands in Meetings/_inbox/ and waits here for triage.
Routing is agent-driven, not script-driven: similarity (via qmd-vector_search) finds where the new note's siblings live; memory (feedback_meeting-routing.md) captures learned exceptions so future runs apply them automatically.
Frontmatter Conventions
Files in _inbox/ are normal meeting notes with Notes: and created: frontmatter. The skill does not modify frontmatter — it only moves files (git mv) and updates the H1 inside when renaming.
Workflow
Step 1: Discover candidates
See vault-context for vault discovery conventions.
- No args: query the live Base view —
obsidian base:query path=Meetings.base view=Inbox format=paths. Returns every file underMeetings/_inbox/(the.gitkeepis excluded automatically since it's not a.md). <path>arg: single-file mode — operate on that path.- If the candidate list is empty, print
Nothing to organize.and exit.
Step 2: Read memory for learned conventions
Check memory/feedback_meeting-routing.md (if present). It contains routing hints the user has confirmed over time, e.g.:
Titles with both "
<name-a>" and "<name-b>" →Meetings/Tech Syncs/(not One-on-ones/), filename<Name A> x <Name B> Sync - YYYY-MM-DD.md.
Apply these before qmd similarity so they always win.
Step 3: For each candidate, propose a destination
For each candidate file:
- Extract the title: filename minus the trailing
- YYYY-MM-DD.md(or use the H1 inside the body if cleaner). - Apply memory overrides (Step 2). If a memory rule matches confidently, skip to Step 4 with
confidence = high (memory). - Otherwise, run qmd-vector_search on the title — return top 5 most similar meeting paths.
- Aggregate:
- Count how many of the 5 matches live in each folder. Pick the mode.
- Inspect filenames in that folder to infer naming pattern:
- All siblings end in
- YYYY-MM-DD.md→ keep date suffix. - All siblings use
YYYY-MM-DD.mdonly (e.g.Engineering/Scrum/) → use date-only. - Siblings share a prefix (e.g.
Tech Leads Weekly -) → consider renaming to match.
- All siblings end in
- Score confidence:
Match count Confidence 5/5 or 4/5 same folder high 3/5 same folder medium < 3/5 or no matches low
Step 4: Present proposals to the user — MANDATORY CONFIRMATION
⚠️ Never move files without confirmation, except in --auto mode (where only high-confidence proposals run, low/medium skip).
Display a numbered table:
| # | File | → Folder | New filename | Confidence | Reasoning |
|---|------|----------|--------------|------------|-----------|
| 1 | _inbox/Jane x Me - 2026-06-02.md | One-on-ones/ | Jane x Me - 2026-06-02.md | high (5/5 qmd) | siblings: John x Me, Alex x Me, Sam x Me |
| 2 | _inbox/PatMe Sync - 2026-06-02.md | Tech Syncs/ | PatMe Sync - 2026-06-02.md | high (memory) | memory: Pat/Me → Tech Syncs |
| 3 | _inbox/Foo Bar - 2026-06-01.md | ? | ? | low (1/5 qmd) | no clear majority — needs user |Ask the user: "all", "high-confidence only", "1,3", "none", or per-file corrections (e.g. "2 → Engineering/").
Step 5: Apply moves
For each approved proposal:
git mv "<old path>" "<new path>"— preserves history.- If the new filename differs from the old, update the H1 inside the file. Skip if H1 already matches the new filename.
Step 6: Capture corrections to memory
When the user overrides a proposal ("don't put Pat/Me in One-on-ones, it goes to Tech Syncs"), this is feedback worth saving. Write or update memory/feedback_meeting-routing.md:
---
name: meeting-routing-conventions
description: Learned routing rules for /organize-meetings — overrides qmd similarity
metadata:
type: feedback
---
When triaging Meetings/_inbox/ via /organize-meetings:
- Titles containing both "<name-a>" and "<name-b>" → `Meetings/Tech Syncs/`.
**Why:** It's a recurring manager sync, not a standard 1:1.
**How to apply:** Apply before qmd similarity. Filename pattern: `<original> - YYYY-MM-DD.md`.
- (more learned rules append here over time)Always include a Why: (the user's reasoning, if given) and How to apply: (the precondition + filename pattern). Capture one rule per bullet so future updates are atomic.
Step 7: Commit
See /commit. Skip when called as part of /meeting wrap pending — the parent sequence commits at the end.
When called standalone: commit with update: /organize-meetings — N files moved (or similar). Skip if nothing was moved.
Edge Cases
- No qmd matches at all (e.g. truly novel meeting type): low confidence. Ask the user; do NOT default to a folder.
- Multiple folders tied: low confidence. Ask the user.
- File already wrapped (
NotesCached,Participants,TodosExtractedalready set in frontmatter): still route it — wrapping doesn't lock the file location. Frontmatter survives the move. - H1 inside the file is generic (e.g. just
# Meeting): leave it alone; don't fabricate a new H1. - Filename collision at destination: append
.dup-Nor ask the user.
Important Notes
- Always use
git mv, nevermv— history matters. - Never auto-move
lowconfidence proposals, even in--automode. - Memory entries are cumulative: append, don't overwrite. Each entry is a separate rule.
- If qmd's index is stale (a recent batch of renames hasn't been reindexed), low-confidence proposals may be common — note this in the proposal reasoning and suggest the user reindex.
- This skill does NOT modify the
Notes:,Participants:,NotesCached:, orTodosExtracted:frontmatter. Those belong to/cache-notes,/fill-participants, and/followup-todos.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

