Meeting Transcript → Action Items

No more re-reading the transcript. No more chasing people to ask "wait, what were you going to do?" The skill reads the full meeting before extracting, so owners and deadlines mentioned anywhere in the conversation are captured correctly.

When to Use / Best For

• You just wrapped a call and need to know who owes what — paste the transcript before you even close the meeting tab
• You have a Zoom, Otter.ai, or Fireflies transcript sitting unread — drop it in and get a task list in seconds
• Your standups produce zero follow-through — extract commitments immediately after each standup and share before everyone forgets
• You took 1:1 notes and need a clean task list for both parties — paste your notes and get each person's items separated out
• You need to share next steps with someone who wasn't in the meeting — the JSON output pastes directly into Notion, Linear, Jira, or any task manager
• You ran a client call and need a professional follow-up summary — extract and share before the client even gets back to their desk

How It Works

Step 1 — Paste your transcript. Any format: Zoom auto-transcription, Otter.ai export, Fireflies summary, Fathom notes, bullet-point meeting notes, or a Slack thread dump.

Step 2 — The skill reads the full meeting first. Before extracting anything, it reads the entire transcript — so an owner mentioned late in the conversation gets correctly assigned to a commitment made earlier.

Step 3 — Every commitment is extracted. Not requests. Not questions. Actual commitments: things people said they would do, or were explicitly assigned. Owner is resolved from context (even "I'll handle it" becomes the right person's name).

Step 4 — You get clean JSON. A structured array with owner, action, and deadline for every item. Deadline is returned exactly as spoken — "before the 15th", "EOD Friday", "next sprint review" — not normalized.

Step 5 — Ask for a table view if you prefer. Say "show me a table" and the same output is formatted for human reading.

Example

Input:

Jordan: I'll send the API spec to the frontend team by Monday — that's still on track.
Sam: I need to look into the rate limit thing we discussed. I'll do that today.
Priya: Sam, can you also loop in the DevOps team about the deployment window we're targeting Friday?
Sam: Yeah I'll ping them this afternoon.
Priya: I'll send the meeting notes out after this call.

Output:

[
  { "owner": "Jordan", "action": "Send API spec to frontend team",   "deadline": "Monday" },
  { "owner": "Sam",    "action": "Investigate rate limit issue",     "deadline": "Today" },
  { "owner": "Sam",    "action": "Ping DevOps about Friday deploy",  "deadline": "This afternoon" },
  { "owner": "Priya",  "action": "Send meeting notes to team",       "deadline": "After this call" }
]

Why This Skill Is Different

• Requests vs. commitments — "can you do X?" only becomes an action item if the person accepts. Unanswered requests are not extracted.
• Pronoun resolution — "I'll take care of it" becomes the speaker's name, not a null owner. Works across speaker-labeled transcripts and unnamed note formats.
• No guessing — if ownership is genuinely ambiguous, owner is returned as null rather than hallucinated. You always know which items need a human decision.
• Any format — timestamped Zoom exports, all-caps no-punctuation auto-transcripts, informal bullet notes, 1:1 manager notes. All handled.

Who It's For

• Managers and team leads who run recurring syncs and need clean handoffs
• Project managers and consultants who bill time and need airtight follow-up records
• Executive assistants who take meeting notes and send action summaries
• Founders and solo operators who jump between calls and can't afford things falling through cracks
• Anyone using Zoom, Otter.ai, Fireflies, or Fathom who wants their transcript to actually produce work