Install
$ agentstack add skill-sidsaladi9-persona-os-meeting-notes ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
Meeting Notes
Meetings end and the value leaks out — decisions get forgotten, action items lose their owners, and the recap never gets sent. This skill takes whatever you have (scribbled notes, a wall-of-text transcript, or a connected transcript source) and turns it into a clean, skimmable summary you can paste into Slack or email the same day. The goal is extraction, not transcription: surface what was decided, who owns what by when, and what's still open.
Grounded in: Death by Meeting — Patrick Lencioni: separate decisions from discussion; every action has an owner and date. Go deeper (The Product Channel): The Product Channel
When to use this
- You have raw notes or a transcript and need a usable summary fast.
- You ran a standup, planning, review, 1:1, or stakeholder sync and owe people a recap.
- You want a clean list of action items with owners and due dates pulled out of the noise.
- Someone asks "what did we actually decide?" and you need a clear answer.
Before you start (gather these)
- The raw notes or transcript. If the user hasn't pasted it, ask them to — or pull it from a connected transcript tool (e.g. Fireflies, Granola) if one is available.
- The meeting type (standup, planning, review, 1:1, stakeholder sync) — it shapes what matters most.
- Who was in the room — needed to assign action item owners by name.
Process
- Read everything first. Don't summarize as you go. Take in the whole input before extracting, so you understand the arc of the conversation and don't mistake early discussion for final decisions.
- Extract DECISIONS — anything the group agreed on or settled. State each as a clear, standalone sentence. If a decision was implied but not confirmed, mark it "(tentative)".
- Extract ACTION ITEMS — every commitment to do something. For each, capture: the action (a clear next step, not a vague topic), the owner (a named person), and the due date. Resolve relative dates (today, Friday, next Tue, EOW) against the meeting date and render them as absolute YYYY-MM-DD — this is resolution, not invention. Write "TBD" only when no date is recoverable — don't invent one.
- Extract OPEN QUESTIONS — anything raised but unresolved, blockers, or items explicitly deferred to a later discussion.
- Capture KEY CONTEXT — the few facts, numbers, or rationale worth keeping that aren't decisions or actions (e.g. "Q3 budget is locked at $40K", "legal flagged the data clause").
- Flag ownerless commitments. If something sounds like a commitment but no one clearly owns it, still list it as an action item with "owner: TBD" and surface it — these are the items that fall through the cracks.
- Stated vs. inferred work. Only surface work that was actually stated. If you want to surface latent or unspoken follow-ups that no one explicitly committed to, that's allowed — but put them under a separate "Suggested follow-ups (not committed)" heading so they're never mistaken for agreed action items.
Output template
> Recap: [Meeting name] — [date] > > TL;DR > - [Line 1: the headline outcome or biggest decision] > - [Line 2: the most important action / who's doing what next] > - [Line 3: anything still open or needing a follow-up] > > Decisions > - [Decision stated as a clear sentence] > - [Decision stated as a clear sentence] (tentative) > > Action items > > | Action | Owner | Due | > |---|---|---| > | [Clear next step] | [Name] | 2026-06-20 | > | [Latent / ownerless item that was flagged] | TBD | TBD | > > Open questions > - [Unresolved item or blocker] > - [Deferred decision] > > Suggested follow-ups (not committed) > - [Inferred next step no one explicitly committed to] > > Notes / context > - [Fact, number, or rationale worth keeping]
Keep it tight and skimmable — the TL;DR alone should be enough for someone who wasn't there to know what happened.
Avoid (anti-patterns)
- Transcribing everything instead of extracting what matters — a summary that's as long as the transcript is useless.
- Action items with no owner or no due date. Aim for both owner and due on every action; use "TBD" rather than leaving a field blank. The exception: flagged ownerless or latent items (Step 6) may legitimately carry "TBD" in both Owner and Due — that's how they get surfaced rather than dropped.
- Burying the decision inside a paragraph of prose. Decisions go in their own list, stated plainly.
- Inventing owners, decisions, or dates that aren't in the source. If it's ambiguous, mark "owner: TBD" or "(tentative)" — don't guess.
Tips
- Assign owners by name, not by role ("Priya" not "the design team") — accountability needs a person.
- Separate decisions from discussion: a thing someone suggested isn't a decision until the group agreed.
- Send the recap the same day, while it's fresh and people can still correct it.
- When in doubt about a due date, write "TBD" and let the owner fill it in — a flagged gap beats a wrong date.
Source & license
This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: Sidsaladi9
- Source: Sidsaladi9/persona-os
- License: MIT
- Homepage: https://sidsaladi.substack.com
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.