meeting-notes-summarizer
Skill0 starsChanged 5 months agoMITView on GitHub
Use when the user provides a meeting transcript, recording transcript, or raw meeting notes and wants a structured summary. Triggers on "summarize this meeting", "extract action items from this transcript", "what did we decide in this call", "TL;DR this Zoom recording", or when the user pastes a transcript longer than ~500 words from a meeting context.
In this repository7 other files it gives its agents.
- api-docs-from-openapiSkill
- changelog-writerSkill
- code-review-preflightSkill
- csv-cleanerSkill
- json-schema-from-sampleSkill
- regex-builderSkill
- sql-from-questionSkill
The file
We split this skill into the rules it gives an agent
6 rules. About 0 are common, 6 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 12
Read the entire input first: Don't summarize as you go.
- Distinctive: line 26
If something was raised but not decided, put it in "Open questions" instead.
- Distinctive: line 34
If the owner wasn't named, write "[unowned]" so it's obvious.
- Distinctive: line 59
If the transcript clearly contains sensitive info (compensation discussions, performance reviews of named individuals, legal disputes), …
- Distinctive: line 63
Transcript with low-quality auto-transcription: if you see frequent garbled phrases ("the [unintelligible] is going to [unintelligible]"), …
- Distinctive: line 64
Transcript with multiple meetings concatenated: ask whether to summarize each separately or combine.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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
Meeting Notes Summarizer
Turn raw transcript into a structured, scannable summary.
Process
- Read the entire input first. Don't summarize as you go. Meeting context shifts; early statements get reversed later.
Output sections in this order:
TL;DR
2-4 sentences. The headline of the meeting. A reader who only reads this should know what the meeting was about and whether it concerns them.
Decisions
Bulleted. Each decision: what was decided, in past-tense, action voice. No hedging.
- "Decided to ship v2 on Friday." OK
- "Discussed shipping v2." FAIL (a discussion isn't a decision)
If something was raised but not decided, put it in "Open questions" instead.
Action items
Bulleted. Each: WHO, WHAT, by WHEN.
- "Alice will publish the API spec by EOD Wednesday."
- "Bob to follow up with legal on the contract; no deadline given."
If the owner wasn't named, write "[unowned]" so it's obvious. If no deadline was set, omit the deadline rather than fabricating one.
Open questions
Bulleted. Things raised that didn't get resolved, with whatever context exists about who's blocking or what's missing.
Discussed (skim-friendly recap)
Brief paragraphs by topic. Use this for stakeholders who want context but didn't attend.
- Tone: factual, neutral, no editorializing. You weren't there. Don't add sentiment ("the team was excited about...") unless the transcript explicitly says it.
- Speaker attribution: include speakers' names when they own a decision or action item. Drop names for general discussion to keep it readable.
What NOT to do
- MUST NOT invent action items. If nobody committed to doing something, don't list it as an action item.
- MUST NOT invent deadlines. If the discussion was vague about timing, the action item is vague about timing.
- MUST NOT include filler: "we then discussed", "the meeting started with", "moving on to the next topic". Strip transcription noise.
- MUST NOT include personal small talk, off-topic asides, or interruptions. Skip them silently.
- MUST NOT quote large chunks verbatim. Summarize. Quote only when a specific phrasing matters (e.g. exact commitment language for a contract conversation).
Privacy considerations
- Mention names that appear in the transcript only in their roles (decisions, action items). Don't add titles, departments, or other identifying info that wasn't said.
- If the transcript clearly contains sensitive info (compensation discussions, performance reviews of named individuals, legal disputes), flag this at the top: "This transcript contains sensitive content. Confirm you want a summary written down before I produce one." Wait for confirmation.
Edge cases
- Transcript with low-quality auto-transcription: if you see frequent garbled phrases ("the [unintelligible] is going to [unintelligible]"), note that the summary may have gaps and suggest the user spot-check.
- Transcript with multiple meetings concatenated: ask whether to summarize each separately or combine.
- Very short transcript (<200 words): the structured format may be overkill. Ask whether they want full structure or just a 1-paragraph summary.
- Transcript in a language other than English: confirm what language to write the summary in. Default to the transcript's language.
Length calibration
- 30-min meeting transcript → ~200-400 word summary
- 1-hour meeting transcript → ~400-700 word summary
- 2+ hours → consider asking if a tiered summary (exec / detailed) is wanted
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.

