agentleFS
serhii12serhii12/vault-os/.claude/skills/ingest-meeting/SKILL.md

ingest-meeting

Skill2 starsChanged 7 months agoMITView on GitHub

Ingest Granola meeting notes (synced into Granola/ by the obsidian-granola-sync plugin) into the LLM Wiki — writes a wiki summary, updates attendee CRM pages, surfaces topic connections and contradictions, drops action items into the right project notes, and logs the ingest with the granola_id so re-runs skip already-processed meetings. Use when the user says "ingest meetings", "/ingest-meeting", "process Granola", "ingest yesterday's meeting", or asks to file new meeting notes into the wiki.

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 2 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read frontmatter (granola_id, title, created, attendees, optional folders) and body sections:

    line 36
  2. Very specific:

    If ## Enhanced Notes is missing (very short meeting or sync error), use Private Notes only and flag low-coverage in the log entry.

    line 42
  3. Distinctive:

    Before drafting writes, search the existing vault:

    line 58
  4. Very specific:

    Create the wiki summary with frontmatter (source: agent · type: wiki-article · topic: meetings · granola: <id>).

    line 109
  5. Very specific:

    Create new files (e.g., People/<NewName>.md from template) only when explicitly confirmed.

    line 111
  6. Very specific:

    Update _agent/wiki/index.md last so counts reflect the writes.

    line 112
  7. Distinctive:

    After writing, print a one-screen summary:

    line 128
  8. Very specific:

    If Granola/ is empty but the user wants to ingest a meeting that already happened, use the Granola MCP (mcp__*Granola*) directly:

    line 148

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Ingest Meeting — Granola → LLM Wiki

Implements Step 1 (Ingest) of _agent/context/knowledge-base-workflow.md for Granola meeting notes. The Granola sync plugin writes the raw layer; this skill compiles it into the wiki.

Conventions

  • Source folder: Granola/ (plugin default). Read-only — see Granola/AGENTS.md.
  • Wiki destination: _agent/wiki/meetings/.
  • Person CRM: People/<Name>.md (existing module). Wiki-layer summaries optional under _agent/wiki/people/.
  • Log: append to _agent/log.md in the parseable ## [YYYY-MM-DD] op | title format with granola_id for dedup.
  • Tag every agent-created file with source: agent per _agent/context/tagging.md.

Workflow

Step 0 — Find new meetings

grep -rl "^type: note$" Granola/ --include="*.md" 2>/dev/null

For each candidate, extract its granola_id from frontmatter and check whether it's already logged:

grep -q "granola_id: <id>" _agent/log.md && echo skip || echo ingest

Already-ingested → skip. Default scope: just process the unlogged ones. If the user passes a date or count, narrow accordingly (--since 2026-05-01, --last 3).

Step 1 — Read each new meeting

Read frontmatter (granola_id, title, created, attendees, optional folders) and body sections:

  • ## Private Notes — user's typed notes (high signal for intent)
  • ## Enhanced Notes — Granola's processed output (decisions, action items, key topics)
  • Optional inline transcript

If ## Enhanced Notes is missing (very short meeting or sync error), use Private Notes only and flag low-coverage in the log entry.

Step 2 — Extract structured data

From the body, identify:

  1. Topics discussed — 2–4 main subjects (project names, problem areas, decisions)
  2. Decisions made — explicit decisions, with rationale if captured
  3. Action items — - [ ]-style or "X will do Y by Z"; capture owner and due date
  4. Open questions — unresolved items
  5. Key insights / quotes — content worth preserving as standalone wiki articles

For each topic, check _agent/wiki/index.md — does it already have a wiki folder? If yes, plan an update; if no and the topic recurs (count mentions across past 30 days of Granola/ files), propose creating it.

Step 3 — Surface connections (mandatory)

Before drafting writes, search the existing vault:

  1. Backlinks for each attendee — grep -rl "\[\[<Name>" --include="*.md" → has the user been talking with them about adjacent topics?
  2. Topic search — for each main topic, grep across 1-Projects/, 2-Areas/, _agent/wiki/, Journal/daily/. Read 1–2 most relevant matches per topic.
  3. Recent daily notes — past 7 days for the same topics; surface "you mentioned this in yesterday's note as well."
  4. Contradictions — does this meeting conflict with _agent/wiki/<topic>/? With a recent daily note? Flag it.
  5. Recurrence — grep -l "<topic>" Granola/*.md → "this is the 3rd meeting on <topic> in 2 weeks; consider creating _agent/wiki/<topic>/_index.md if it doesn't exist."

Step 4 — Draft writes (do NOT write yet)

Build a single confirmation block listing every proposed write, grouped by destination:

## Proposed writes for "<Meeting title>" (granola_id: <id>)

### Wiki summary (new)
_agent/wiki/meetings/<YYYY-MM-DD>-<slug>.md
  — sections: Attendees · Topics · Decisions · Action items · Open questions
  — links back to: [[Granola/<title>]]

### People (CRM updates)
People/<Name1>.md
  — append to "Recent interactions": <date> — <one-line>
  — owed: <action items they own>
People/<Name2>.md  (NEW — confirm scaffold?)
  — create from _admin/templates/para/Person.md

### Topics
_agent/wiki/<topic>/_index.md
  — append to "Sources": [[meetings/<slug>]]
  — flag contradiction with [[<existing-article>]]: <one-line>

### Action items → tasks
1-Projects/<project>/<file>.md
  — append: - [ ] <task> 📅 2026-05-12 (assigned: <Name>)
Journal/daily/2026-05-06.md
  — append to Tasks: - [ ] <task> 📅 2026-05-08

### Index + Log
_agent/wiki/index.md  — bump meetings topic counts
_agent/log.md          — ## [<date>] ingest | Granola: <title> (granola_id: <id>)

Step 5 — Confirm and apply

Print the block above and ask once:

Apply all proposed writes? y = all · n = abort · s = let me select

On s, walk each destination and ask y/n. On y, apply atomically:

  1. Create the wiki summary with frontmatter (source: agent · type: wiki-article · topic: meetings · granola: <id>).
  2. Append to existing files (people, topic indexes, project notes, daily notes).
  3. Create new files (e.g., People/<NewName>.md from template) only when explicitly confirmed.
  4. Update _agent/wiki/index.md last so counts reflect the writes.
  5. Append to _agent/log.md last:

    ## [2026-05-06] ingest | Granola: Q2 planning sync (granola_id: abc123)
    - source: Granola/2026-05-06 - Q2 planning sync.md
    - wiki: _agent/wiki/meetings/2026-05-06-q2-planning-sync.md
    - touched: People/Alex, People/Sam, _agent/wiki/projects/q2-roadmap, 1-Projects/Q2-Roadmap.md
    - actions: 3 task(s) filed
    - granola_id: abc123

    The granola_id line is the dedup key — Step 0 of the next run greps for it.

Step 6 — Report

After writing, print a one-screen summary:

Ingested 2 meetings.
- Q2 planning sync (45m, 4 attendees) → 3 tasks, 2 topic updates, 1 contradiction flagged
- 1:1 with Alex (30m) → 1 task, 1 open question filed

Look at: _agent/wiki/meetings/2026-05-06-q2-planning-sync.md (contradiction with [[product-strategy]])

Style + safety

  • Never modify files in Granola/ — see Granola/AGENTS.md.
  • Never write to People/, 1-Projects/, or Journal/daily/ without confirmation — those are human territory.
  • Skip empty sections in the wiki summary rather than padding with "None".
  • Cite the source in every wiki article: link back to [[Granola/<file>]].
  • Flag low-coverage ingests in the log so the user knows which meetings have thin content.

Fallback when the plugin isn't running

If Granola/ is empty but the user wants to ingest a meeting that already happened, use the Granola MCP (mcp__*Granola*) directly:

  1. mcp__*Granola*list_meetings for the target date.
  2. mcp__*Granola*get_meeting_transcript for the chosen one.
  3. Synthesize a Granola/<date> - <title>.md file matching the plugin's schema (frontmatter + Private Notes + Enhanced Notes) so subsequent runs find it normally.
  4. Continue from Step 2.

When to skip

  • File has type: transcript — only ingest the matching type: note. Transcripts get linked from the wiki summary but aren't ingested separately.
  • Meeting < 5 minutes with no Enhanced Notes — log as skipped: thin content and move on.
  • granola_id already in _agent/log.md — skip silently unless the user passes --reingest.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.