meeting-notes-to-actions
Skill2 starsChanged 4 months agoMITView on GitHub
Converts raw meeting transcripts, call recordings notes or messy bullet points into a structured summary with decisions, owned action items, deadlines, open questions and risks, separating what was agreed from what was merely discussed. Use when the user shares meeting notes, a transcript, standup or retro output, interview notes or a call recap and wants a summary, follow-ups, action items or a status update derived from it.
9 other projects use a version of this.
Adopted by
Most recently taken up by
- Lethe044/agentskill-kittheir own version, first seen Nov 1, 2026
- Amazon-Quick/Amazon-Quick-official-catalogtheir own version, first seen Oct 28, 2026
- dss-time/repo-doctor-skillstheir own version, first seen Oct 15, 2026
- Baba01hacker666/dummyskillstheir own version, first seen Oct 13, 2026
- SkillMedev/remote-team-toolkittheir own version, first seen Oct 5, 2026
- create-aquarius/business-skills-jatheir own version, first seen Sep 17, 2026
Other versions9 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 9
- Amazon-Quick/Amazon-Quick-official-catalog★ 50same name
- Lethe044/agentskill-kit★ 48same name
- rickbauer327-cell/solo-stack-starter★ 2same name
- ahnchen1983/agentic-substrate★ 1same name
- dss-time/repo-doctor-skills★ 1same name
- Baba01hacker666/dummyskills★ 0same name
- create-aquarius/business-skills-ja★ 0same name
- kakarot-oncloud/claude-skills★ 0same name
- SkillMedev/remote-team-toolkit★ 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 repository13 other files it gives its agents.
- AGENTS.mdAGENTS.md
- accessibility-auditorSkill
- contract-reviewerSkill
- csv-data-profilerSkill
- dependency-auditorSkill
- git-workflow-assistantSkill
- prompt-engineering-optimizerSkill
- release-notes-generatorSkill
- rest-api-designerSkill
- security-code-reviewSkill
- sql-query-optimizerSkill
- technical-writing-editorSkill
- web-research-synthesizerSkill
The file
We split this skill into the rules it gives an agent
2 rules. 0 are common, 2 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 7 edited versions we read keep these.
2
Distinctive
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 24
Not for: summarising documents (no speakers, no commitments) or drafting an agenda for a *future* meeting.
- Distinctive: line 43
Never promote an idea to an action.
“Common” is measured: kept by at least 6 in 10 of the 7 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 Notes → Actions
Extract commitments from conversation. Attribute nothing that was not said.
When to use
Transcripts, standups, retros, planning sessions, client calls, 1:1s, interview debriefs, incident reviews.
Not for: summarising documents (no speakers, no commitments) or drafting an agenda for a future meeting.
Core discipline
The single biggest failure mode is inventing certainty. A conversation is full of half-commitments. Classify each item honestly:
| Signal in the transcript | Classification |
| "We'll go with X", "Decided: X", "Agreed" | Decision |
| "I'll do X by Friday" | Action — owner + date |
| "Someone should look at X" | Action — owner UNASSIGNED, flag it |
| "Maybe we could X", "What if we X" | Idea — not an action |
| "I think X is true" (unverified) | Open question |
| "That'll break if Y happens" | Risk |
| "We can't ship until legal replies" | Blocker / dependency |
| Disagreement with no resolution | Unresolved — record both positions |
Never promote an idea to an action. Never invent an owner. Never invent a date.
Workflow
1. Pass one — map the meeting
Identify participants and roles, the meeting's purpose, and the topic boundaries. Note timestamps or line references if present; you will cite them.
1b. Run the extractor
python3 scripts/extract_actions.py transcript.txt --meeting-date 2026-07-28It splits speaker turns, classifies each sentence as commitment / decision / idea / question / risk / parked / reversal, attributes a speaker, resolves relative dates against the meeting date, and marks every unowned or undated item.
It deliberately over-collects. Its output is a candidate list you verify in pass two — never paste it into the final notes unreviewed.
2. Pass two — extract, with evidence
For every candidate item, record the supporting quote. If you cannot quote it, it does not go in the output.
Action items need five fields:
| Field | Rule |
| What | One verb-first sentence, concretely completable |
| Owner | A named person. UNASSIGNED if never stated — never guess from role |
| Due | Absolute date. Resolve "next Friday" against the meeting date and mark it (inferred). NO DATE if none |
| Dependency | What must happen first, if stated |
| Evidence | The quote or line reference |
3. Pass three — resolve references
- Relative dates → absolute, using the meeting date; mark inference.
- Pronouns → named people; if genuinely ambiguous, write
[AMBIGUOUS: "he" — could be Marc or Tom]rather than choosing. - Jargon and acronyms → expand on first use if inferable from context, else flag.
- Duplicates → merge, keeping the strongest commitment.
- Superseded items → keep only the final version, and note the reversal if the change matters ("initially agreed X, reversed to Y at 34:10").
4. Pass four — audit before writing
- Does every action have a quote behind it?
- Any action with an owner who never spoke a commitment? Downgrade it.
- Any decision that was actually still being debated at the end? Move it to Unresolved.
- Anything sensitive (compensation, personnel, legal exposure)? Flag it for restricted distribution rather than dropping it silently.
Output format
# <Meeting title> — <YYYY-MM-DD>
**Attendees:** … **Duration:** … **Purpose:** one line
## TL;DR
Three bullets maximum. What changed as a result of this meeting.
## Decisions
| # | Decision | Rationale | Decided by | Reversible? |
## Action items
| # | Action | Owner | Due | Depends on | Evidence |
## Open questions
| Question | Who can answer | Needed by |
## Risks & blockers
| Item | Impact | Mitigation discussed | Owner |
## Discussed, not decided
Bullets — ideas parked without commitment.
## Flags
- Unassigned actions: N
- Actions without dates: N
- Ambiguities: …Order sections by what the reader acts on first. If the meeting produced no decisions, say so explicitly — that is useful information.
Tools
| Script | Use |
scripts/extract_actions.py | Candidate extraction with speaker attribution and date resolution |
assets/meeting-notes.md | Output template matching the format above |
python3 scripts/extract_actions.py notes.txt --meeting-date 2026-07-28 --format mdThe script resolves "next Tuesday", "end of week", "in 3 weeks" and "Q4" to absolute dates and always labels them as inferred. It flags "asap"/"soon" as non-dates rather than guessing.
Worked example
Input
Sarah: …so the migration is behind. We said end of Q3 but realistically that's not happening. Marc: What's the blocker? Sarah: Mainly the schema rewrite. Tom's been pulled onto the billing incident. Marc: Okay. I think we push to Q4. Anyone object? … Right, Q4 it is. I'll update the roadmap deck before the board meeting. Sarah: I can take the schema rewrite off Tom's plate, get you a plan by next Tuesday. Marc: Do we need to tell the Acme account team? Sarah: Probably? Their integration depends on it. Someone should check the contract for a date commitment. Marc: Let's park that until we know the Q4 date is firm.
Output (extract) — meeting date 2026-07-28
Decisions
| # | Decision | Rationale | Decided by | Reversible? |
| D1 | Migration deadline moves from end of Q3 to Q4 | Schema rewrite blocked; Tom reassigned to the billing incident | Marc, no objections raised | Yes |
Action items
| # | Action | Owner | Due | Depends on | Evidence |
| A1 | Update the roadmap deck with the Q4 migration date | Marc | Before the board meeting NO DATE — board date not stated | D1 | "I'll update the roadmap deck before the board meeting" |
| A2 | Take over the schema rewrite from Tom and deliver a plan | Sarah | 2026-08-04 (inferred from "next Tuesday") | — | "get you a plan by next Tuesday" |
Open questions
| Question | Who can answer | Needed by |
| Does the Acme contract commit us to a migration date? | Legal / account team | Once the Q4 date is firm |
| Must the Acme account team be notified, and when? | Marc | Deferred — explicitly parked |
Discussed, not decided
- Notifying the Acme account team — parked pending confirmation of Q4.
Flags
- "Someone should check the contract" — no owner assigned; needs one.
- A1 has no concrete due date; the board meeting date was never stated.
Quality checklist
extract_actions.pyoutput reviewed candidate by candidate, not pasted through- Every action traceable to a quote
- No owner assigned who did not accept the work
- Every relative date resolved and marked
(inferred) - Ideas kept out of the action list
- Unresolved disagreements recorded, not smoothed over
- Reversals and superseded decisions handled
- Unassigned actions and missing dates surfaced in Flags
- TL;DR readable in 15 seconds
- Sensitive content flagged for restricted distribution
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.

