Meeting To Tasks

Meeting to Tasks

Transform raw meeting transcripts into structured, actionable outputs: decisions, action items, owners, deadlines, and open questions -- including implicit commitments participants may not realize they made.

Contents

• references/extraction-schema.md -- YAML field schemas for every extraction phase
• references/markers.md -- Decision/action/implicit markers, priority + deadline inference, confidence scoring
• references/output-templates.md -- Meeting summary, task files (Linear/GitHub/generic), follow-up email, directory layout
• references/meeting-types.md -- Meeting-type detection and per-type focus

Workflow

1. Accept input. Handle plain text, Markdown, pasted text, audio transcript exports (Otter.ai, Fireflies, Rev, Zoom, Google Meet, Microsoft Teams), or structured notes. 2. Validate. Confirm there is enough content (~100 words minimum), identifiable speakers, a discernible topic, and any timestamps. If input is too sparse, ask for the full transcript, the participant list, or the meeting purpose before proceeding. 3. Read the full transcript first. Do not extract from a partial read -- later context can change earlier interpretation. 4. Detect the meeting type and adjust focus per references/meeting-types.md. 5. Extract in phases using the schemas in references/extraction-schema.md: metadata, decisions, action items, open questions, parking lot, discussion points. Apply the markers and inference rules in references/markers.md. Preserve the exact source quote for every item. 6. Identify speakers. Build the participant list from speaker labels and track who said what. When labels are absent, infer from context clues, otherwise mark "Unassigned" and flag for the user to assign. 7. Score confidence on every item (high/medium/low). Flag all low-confidence items for human review per references/markers.md. 8. Resolve conflicts and ambiguity: flag contradictory commitments as "CONFLICT" with both versions; list all candidates for unclear ownership; translate vague deadlines ("soon", "ASAP") to specific dates and flag for confirmation; offer both readings of ambiguous scope; consolidate duplicates. 9. Generate outputs using references/output-templates.md. Confirm which project management format(s) the user wants before generating task files. Always generate the follow-up email, even if unasked. 10. Post-process: deduplicate overlapping items; map dependencies; highlight the critical path; report owner imbalance; warn on deadline clustering; list items missing owners or deadlines.

Rules

• Make action items actionable: start each with a verb, specific enough that the owner knows exactly what to do. Transform "Think about pricing" into "Research competitor pricing and propose new pricing tiers by [date]."
• Err toward capturing too much; a flagged low-confidence item beats a missed commitment.
• Do not editorialize. Report what was said; keep any analysis in clearly labeled "Analysis" sections.
• Respect privacy. If the meeting contains HR issues, personal matters, or confidential information, flag it and ask how to handle it in outputs.
• Handle multi-language transcripts: extract in the primary language and note language-specific nuances.
• Maintain an extraction log noting confidence levels, ambiguities, and decisions made during extraction.

Quick Commands

• "Extract from [file]" -- Full extraction from the specified transcript
• "Just action items" -- Action items only (skip decisions, questions, summary)
• "Generate tasks for Linear" / "Generate tasks for GitHub" -- Action items in that format
• "Draft follow-up email" -- Follow-up email only
• "Who owes what?" -- Owner-grouped view of all action items
• "What was decided?" -- Decisions only
• "What is still open?" -- Open questions and unresolved items

Extraction Schema

Structured fields to capture during each extraction phase. Always preserve the exact source quote that produced each item.

Phase 1: Meeting Metadata

meeting:
  title: ""                # Inferred from content or asked
  date: ""                 # Extracted or today's date
  duration: ""             # If available
  type: ""                 # standup, planning, review, brainstorm, 1:1, all-hands, client, etc.
  participants:
    - name: ""
      role: ""             # Inferred from context (facilitator, presenter, decision-maker, etc.)
  agenda_items: []         # If an agenda was referenced
  context: ""              # Brief description of what the meeting was about

Phase 2: Decisions

decisions:
  - id: "D-001"
    decision: ""           # Clear statement of what was decided
    context: ""            # What led to this decision
    made_by: ""            # Who made or confirmed the decision
    participants: []       # Who was present for the decision
    confidence: ""         # high, medium, low (how clearly was this a decision?)
    conditions: ""         # Any conditions or caveats
    reversibility: ""      # easy, moderate, hard (can this be undone?)
    source_quote: ""       # Exact quote from transcript

Phase 3: Action Items

action_items:
  - id: "A-001"
    title: ""              # Clear, concise action statement (imperative mood)
    description: ""        # Detailed description with context
    owner: ""              # Person responsible
    collaborators: []      # Others involved
    deadline: ""           # Explicit deadline, inferred deadline, or "TBD"
    priority: ""           # high, medium, low (inferred from urgency signals)
    status: "open"         # open, in-progress, blocked, done
    type: ""               # task, research, decision-needed, follow-up, communication
    dependencies: []       # Other action items this depends on
    related_decisions: []  # Decision IDs this action relates to
    commitment_type: ""    # explicit, implicit, inferred
    source_quote: ""       # Exact quote from transcript
    confidence: ""         # high, medium, low (how certain is it this is an action item?)
    project: ""            # Which project or workstream this belongs to
    tags: []               # Categorization tags

Phase 4: Open Questions

open_questions:
  - id: "Q-001"
    question: ""           # The question as stated
    raised_by: ""          # Who asked
    context: ""            # Why it matters
    assigned_to: ""        # Who should answer (if identified)
    deadline: ""           # When an answer is needed
    related_items: []      # Related decisions or action items
    source_quote: ""

Phase 5: Parking Lot

parking_lot:
  - id: "P-001"
    topic: ""
    raised_by: ""
    reason_deferred: ""    # Why it was not addressed
    follow_up_meeting: ""  # When it should be revisited

Phase 6: Key Discussion Points

discussion_points:
  - topic: ""
    summary: ""            # 2-3 sentence summary
    participants: []       # Who contributed
    outcome: ""            # decision, action, deferred, informational
    time_spent: ""         # estimated, if timestamps available

Detection Markers and Inference Rules

Lookup tables for identifying decisions, action items, and implicit commitments, plus the rules for inferring priority and deadlines.

Decision Markers

Explicit decision markers:

• "We decided to..."
• "Let's go with..."
• "The decision is..."
• "We're going to..."
• "We agreed that..."
• "Final answer is..."
• "Approved."
• "Let's move forward with..."
• "That's the plan."

Implicit decision markers:

• Consensus after debate: multiple people agreeing on a direction after discussion
• Leader pronouncement: a senior person stating direction without objection
• Default by silence: proposal made with no objection raised
• Conditional decision: "If X, then we'll do Y"

Action Item Markers

Explicit action item markers:

• "I will..."
• "Can you [do X]?"
• "Let's [do X] by [date]"
• "[Name] will..."
• "Action item: ..."
• "TODO: ..."
• "Next step is..."
• "We need to..."
• "Make sure to..."
• "Follow up on..."
• "Schedule a..."
• "Send [something] to [someone]"
• "Update [something]"
• "Review [something]"
• "Prepare [something]"
• "Set up [something]"

Implicit commitment markers (critical -- people often do not realize they committed):

• "I'll look into that" = ACTION: Research and report back
• "Let me check" = ACTION: Investigate and share findings
• "I can probably..." = ACTION: Attempt and confirm
• "We should..." (when said by someone with authority) = ACTION: Do it
• "That's a good point, let me think about it" = ACTION: Consider and respond
• "I'll circle back on that" = ACTION: Follow up
• "Let me talk to [person]" = ACTION: Have conversation and report back
• "I think we need to..." (with group agreement) = ACTION: Initiate
• "Yeah, I can do that" = ACTION: Explicit acceptance of task
• "I'll take care of it" = ACTION: Ownership accepted
• "We'll figure it out" = ACTION: Needs decomposition into specific tasks
• "I'll ping you" = ACTION: Send follow-up communication
• "Let me get you that" = ACTION: Provide deliverable

Priority Inference Rules

• HIGH: Mentioned as urgent, blocking other work, has a tight deadline, or requested by a senior leader
• MEDIUM: Standard action item with a deadline, part of normal workflow
• LOW: Nice-to-have, exploratory, no deadline mentioned

Deadline Inference Rules

• Explicit: "by Friday", "before the next meeting", "end of week"
• Inferred from context: "before the launch" (if launch date is known), "before the next sprint"
• Meeting cadence: If it is a weekly meeting, default deadline is "before next meeting" (1 week)
• If no deadline signal at all: Mark as "TBD" and flag for owner to set

Confidence Scoring

• HIGH: Clear, explicit statement. Speaker named. Unambiguous meaning.
• MEDIUM: Reasonable inference from context. Speaker identifiable. Meaning is likely correct but could be interpreted differently.
• LOW: Implicit commitment or ambiguous statement. Speaker may be unclear. Flag for human review.

For LOW confidence items, always flag them for review:

[LOW CONFIDENCE] The following items were extracted but may not be accurate.
Please review and confirm or remove:

- A-XXX: [Action] - Assigned to [Name]
  Reason for low confidence: [Explanation]
  Original quote: "[Quote]"

Meeting Type Detection

Detect the meeting type from content and adjust extraction focus and output emphasis accordingly.

Standup / Daily Sync

Focus on: Blockers, what was done yesterday, what is planned today. Output emphasis: Blockers list, brief status summary.

Sprint Planning / Backlog Grooming

Focus on: Stories accepted, estimates, sprint commitments. Output emphasis: Sprint backlog with story points, capacity allocation.

Retrospective

Focus on: What went well, what did not, improvements. Output emphasis: Improvement action items with owners.

1:1 Meeting

Focus on: Career development, feedback, personal action items. Output emphasis: Private action items, feedback themes (handle sensitively).

Client Meeting

Focus on: Requirements, commitments to client, follow-up items. Output emphasis: Client-facing follow-up email, internal action items separately.

All-Hands / Town Hall

Focus on: Announcements, Q&A answers, organizational decisions. Output emphasis: Key announcements summary, FAQ compilation.

Brainstorm / Workshop

Focus on: Ideas generated, ideas selected, next steps for exploration. Output emphasis: Idea catalog with prioritization, exploration assignments.

Review / Demo

Focus on: Feedback received, approval decisions, revision requests. Output emphasis: Feedback items with priority, revision task list.

Output Templates

Templates for every generated artifact and the directory layout that holds them.

Output Directory Structure

meeting-outputs/
  {date}-{meeting-slug}/
    meeting-summary.md           # Complete meeting summary
    follow-up-email.md           # Ready-to-send follow-up
    action-items.yaml            # All action items in structured format
    tasks/
      linear/                    # Linear-compatible task files
        A-001.md
        A-002.md
      github/                    # GitHub Issues-compatible files
        A-001.md
        A-002.md
      generic/                   # Generic YAML task files
        A-001.yaml
        A-002.yaml
    raw/
      transcript.md              # Original transcript (preserved)
      extraction-log.md          # Log of extraction decisions and confidence scores

Output 1: Meeting Summary (meeting-summary.md)

# Meeting Summary: [Title]
**Date**: [Date]
**Duration**: [Duration]
**Participants**: [List]
**Type**: [Meeting type]

---

## Summary

[2-3 paragraph narrative summary of the meeting. What was discussed, what was decided,
and what needs to happen next. Written for someone who was not in the meeting.]

---

## Decisions Made

| # | Decision | Made By | Confidence |
|---|----------|---------|------------|
| D-001 | [Decision statement] | [Name] | [High/Medium/Low] |

---

## Action Items

| # | Action | Owner | Deadline | Priority | Type |
|---|--------|-------|----------|----------|------|
| A-001 | [Action statement] | [Name] | [Date] | [H/M/L] | [Type] |

### Implicit Commitments Detected

These items were not explicitly called out as action items but represent commitments made during the meeting:

| # | Commitment | Who Said It | Original Quote |
|---|-----------|-------------|----------------|
| A-XXX | [Interpreted action] | [Name] | "[exact quote]" |

---

## Open Questions

| # | Question | Raised By | Assigned To | Deadline |
|---|----------|-----------|-------------|----------|
| Q-001 | [Question] | [Name] | [Name] | [Date] |

---

## Parking Lot

| Topic | Raised By | Follow-Up |
|-------|-----------|-----------|
| [Topic] | [Name] | [When/Where] |

---

## Discussion Notes

### [Topic 1]
[Summary of discussion]

### [Topic 2]
[Summary of discussion]

Output 2: Individual Task Files

Generate one task file per action item in a format compatible with the user's project management tool.

Linear-Compatible Format (tasks/linear/)

---
title: "[Action item title]"
assignee: "[Owner name]"
priority: "[urgent|high|medium|low]"
status: "Todo"
labels: ["meeting-action", "[project]"]
due_date: "[YYYY-MM-DD]"
---

## Description

[Detailed description of the task]

## Context

This action item came from the [Meeting Title] meeting on [Date].

**Original quote**: "[Source quote from transcript]"

**Related decisions**: [List any related decisions]

## Acceptance Criteria

- [ ] [Specific criterion 1]
- [ ] [Specific criterion 2]

## Dependencies

- [List dependencies]

GitHub Issues Format (tasks/github/)

---
title: "[Action item title]"
assignees: ["[github-username]"]
labels: ["meeting-action", "[priority]"]
milestone: "[if applicable]"
---

## Description

[Detailed description]

## Context

From: [Meeting Title] ([Date])
Owner: [Name]
Deadline: [Date]
Priority: [Priority]

## Tasks

- [ ] [Subtask 1]
- [ ] [Subtask 2]

## Related

- Decision: [Related decision]
- Meeting: [Link to meeting summary]

Generic Task Format (tasks/generic/)

task:
  id: "A-001"
  title: ""
  description: ""
  owner: ""
  deadline: ""
  priority: ""
  status: "open"
  tags: []
  subtasks: []
  notes: ""
  source: "Meeting: [Title] on [Date]"

Output 3: Follow-Up Email Draft (follow-up-email.md)

Subject: [Meeting Title] - Summary and Action Items ([Date])

Hi team,

Thank you for the productive meeting today. Here is a summary of what we covered,
the decisions we made, and the action items with owners and deadlines.

## Key Decisions
[Numbered list of decisions]

## Action Items
[Table of action items with owners and deadlines]

## Open Questions
[List of questions that still need answers]

## Next Meeting
[Date/time of next meeting, if known]

Please review the action items assigned to you and let me know if any deadlines
need to be adjusted.

Best,
[Name]