meeting-summary
Skill1 starsChanged 4 months agoMITView on GitHub
Use when the user provides a meeting reference — a SharePoint recording URL, a Teams meeting/join link, or a meeting named by title or date — and wants it summarized, or asks to "summarize the meeting", extract decisions, or pull action items from a recorded meeting. Fetches the calendar event and transcript via the Microsoft 365 MCP tools and writes structured markdown notes (one meeting note plus per-project docs) into the meetings notes repo, organized by meeting series then date. Transcript-grounded only — never summarizes a meeting it could not fetch a transcript for.
10 other projects use a version of this.
Adopted by
Most recently taken up by
- GreatFigure8/Claude-Skillstheir own version, first seen Oct 25, 2026
- ngduydu/AI-SKILLStheir own version, first seen Oct 13, 2026
- yannick-bottino/consulting-claude-skillstheir own version, first seen Sep 21, 2026
- bam-bam-2/solo-skillstheir own version, first seen Sep 3, 2026
- DGA-SD/Claude-SKILLtheir own version, first seen Aug 15, 2026
- wl-open-source/claude-skillstheir own version, first seen Jul 15, 2026
Other versions10 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 10
- bam-bam-2/solo-skills★ 366same name
- geegewu/gegeewu-open-skill★ 8same name
- drunkcoding/AgentSkillsArxiv★ 5same name
- qike7777/skill-library★ 1same name
- yannick-bottino/consulting-claude-skills★ 1same name
- DGA-SD/Claude-SKILL★ 0same name
- GreatFigure8/Claude-Skills★ 0same name
- ngduydu/AI-SKILLS★ 0same name
- wl-open-source/claude-skills★ 0same name
- XiaMuYao/claude-skill★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository21 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- ai-commentsSkill
- ai-draftSkill
- ai-nextSkill
- ai-postSkill
- call-traceSkill
- draft-prSkill
- dto-mappingSkill
- handoffSkill
- hum-commentsSkill
- jira-dcSkill
- plan-from-issueSkill
- plan-nextSkill
- plan-reviewSkill
- plan-saveSkill
- pr-commentsSkill
- pr-statsSkill
- review-sessionSkill
- skill-builderSkill
- validator-generatorSkill
- templateSkill
The file
We split this skill into the rules it gives an agent
9 rules. 0 are common, 5 are distinctive, and 4 are very specific to this project.
0
Common
Most of the 4 edited versions we read keep these.
5
Distinctive
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 34
Before replying or asking anything, resolve the notes root, silently:
- Very specific: line 36
Read ~/.claude/meeting-summary.local (i.e.
- Very specific: line 40
Setup: Write this template to ~/.claude/meeting-summary.local (never clobber a real value), tell the user to fill in the path to their …
- Very specific: line 89
Resolve the user's identity: Call get_me once — the returned name/email defines whose action items go under "My Follow-ups".
- Distinctive: line 96
Use them to separate what's NEW from what was already decided, and to spot prior open items this meeting resolved.
- Very specific: line 105
Write the notes: Read references/note-formats.md for the exact file paths, naming, and markdown templates, then write the meeting note and …
- Distinctive: line 113
Report and stop: Use the Output format below.
- Distinctive: line 131
Full transcript, always: Never truncate or sample a long transcript.
- Distinctive: line 140
NEVER fabricate a summary without a transcript: No transcript means a clearly-marked skeleton (with consent) or nothing.
“Common” is measured: kept by at least 6 in 10 of the 4 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Meeting Summary: turn a recorded Teams meeting into structured notes
You are turning one recorded Teams meeting into durable markdown notes: a meeting note filed under its series folder, plus a per-project doc for each workstream it touched. Everything you write must be grounded in the fetched transcript — the cardinal failure here is fabricating a summary when the transcript is missing or partial, or silently overwriting a note that already exists. If you can't fetch the transcript, say so and stop; if the note exists, ask.
First action on every invocation — the config gate
Before replying or asking anything, resolve the notes root, silently:
- Read
~/.claude/meeting-summary.local(i.e.$HOME/.claude/meeting-summary.local). - A line
NOTES_ROOT="<path>"with a real existing directory → proceed. - File missing, value empty, or still a
<...>placeholder → run Setup, then stop.
Setup: Write this template to ~/.claude/meeting-summary.local (never clobber
a real value), tell the user to fill in the path to their clone of the meetings
notes repo and save, then stop:
# meeting-summary config — per-user, never committed.
# Set NOTES_ROOT to your local clone of the meetings notes repo.
NOTES_ROOT="<PATH_TO_YOUR_MEETINGS_NOTES_REPO>"When to use this skill
- User pastes a SharePoint recording link (
https://...sharepoint.com/.../Recordings/...) - User pastes a Teams meeting or join link
- User asks to summarize a meeting by name or date ("summarize this morning's sprint review")
- User wants decisions or action items extracted from a recorded meeting
When NOT to use this skill
- Summarizing an email thread, chat, or document — that's ad-hoc summarization, not this skill; no calendar event or transcript is involved.
- Live/future meetings — there is no transcript yet; say so and stop.
- Editing existing notes by hand ("fix the attendee list in Tuesday's note") — just edit the file directly.
Steps
- Identify the meeting. Locate the calendar event with
outlook_calendar_search:- SharePoint recording link: URL-decode the filename; extract the meeting
name (text before the
-YYYYMMDD_HHMMSS-Meeting Recording.mp4suffix) and the date. Search by name, narrow by date. - Teams join link: search by any keywords in the link, or list events near
the implied date and match on the event's
onlineMeetingjoin URL. - Name/date: search directly. Multiple plausible matches → present them via AskUserQuestion and stop until answered. Zero matches → report what you searched and stop.
- SharePoint recording link: URL-decode the filename; extract the meeting
name (text before the
- Read the calendar event.
read_resourcewithcalendar:///events/{eventId}. Record: subject, series name (recurring-series subject; for a one-off, the subject itself), start/end + timezone, attendees,meetingTranscriptUrl, and any agenda in the body. - Fetch the transcript.
read_resourcewith the event'smeetingTranscriptUrlverbatim (formatmeeting-transcript:///events/{token}). For a recurring series, append?start={iso}&end={iso}to scope to the occurrence. Delivered in chunks → fetch all chunks before summarizing. No transcript → do not summarize. Tell the user and offer a skeleton note (metadata only, clearly marked as untranscribed) via AskUserQuestion; only write it if they say yes. - Resolve the user's identity. Call
get_meonce — the returned name/email defines whose action items go under "My Follow-ups". Never assume an identity from the notes or config. - Gather prior context. Glob
{NOTES_ROOT}/Projects/*.mdand the meeting's series folder{NOTES_ROOT}/{Series Name}/*.md; read the docs whose names match projects, products, or prior meetings referenced in the transcript. Use them to separate what's NEW from what was already decided, and to spot prior open items this meeting resolved. - Analyze the transcript. Extract: discussion topics (grouped, each with a concise summary), decisions (stated precisely, with who confirmed them), team follow-ups (owner + deadline when stated), the user's own follow-ups (explicit assignments and clear volunteerings), the distinct projects/workstreams discussed, and any callbacks to prior meetings.
- Write the notes. Read
references/note-formats.mdfor the exact file paths, naming, and markdown templates, then write the meeting note and create/update one project doc per project identified. If the meeting note already exists at the target path, ask (overwrite / skip) via AskUserQuestion before touching it. Project docs are append-only history — add the new entry at the top of Meeting History, check off resolved items, never rewrite prior entries. - Report and stop. Use the Output format below. Do not commit or push the notes repo — writing the files is where this skill ends.
Output format
Meeting note: {NOTES_ROOT}\{Series Name}\{YYYY-MM-DD} {Meeting Name}.md
Project docs: {created: X, updated: Y — names}
Summary: {topics} topics, {decisions} decisions, {follow-ups} follow-ups ({n} yours)
Resolved from previous meetings: {list, or "none"}Rules
What to do
- Ground every claim in the transcript. Decisions verbatim-precise with the decider named; follow-ups only when actually assigned or volunteered.
- Full transcript, always. Never truncate or sample a long transcript; fetch every chunk first.
- Ask at every fork. Ambiguous meeting match, recurring occurrence, existing note at the target path, unclear project grouping → AskUserQuestion, never guess.
- Identity comes from
get_me, paths come from the config file. Nothing personal is hardcoded in this skill.
What NOT to do
- NEVER fabricate a summary without a transcript. No transcript means a clearly-marked skeleton (with consent) or nothing.
- NEVER overwrite an existing meeting note unprompted, and never rewrite or delete prior entries in a project doc.
- NEVER commit or push the notes repo — file writes only.
- Don't paraphrase decisions loosely — a softened decision is a wrong decision.
Format discipline
- The deliverable is the note files plus the four-line report. No transcript replay in chat, no narrating each step, no recap of the note contents beyond the counts.
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.

