| name | meeting-distill |
| description | Extract long-term knowledge from a meeting transcript and persist it into docs/. Filters out ephemeral content and focuses on durable architectural, product, domain, and team knowledge. |
| user_invocable | true |
| allowed-tools | Read, Grep, Glob, Edit, Write, Agent, Bash(git log *), Bash(git diff *) |
| argument-hint | [paste transcript or path to transcript file] |
Meeting Distill
Extract long-term relevant knowledge from a meeting transcript and persist it into the docs/ folder. This skill is designed for recurring team syncs, all-hands, and planning meetings.
Argument: $ARGUMENTS
Step 1: Obtain the transcript
- If
$ARGUMENTS contains a file path, read it.
- If
$ARGUMENTS is empty or short, the transcript was likely pasted directly in the conversation. Use the conversation context.
- If no transcript is available, ask the user to paste one or provide a file path.
Step 2: Identify participants and meeting type
Scan the transcript and note:
- Who attended (names and roles, if discernible)
- Meeting type (weekly sync, all-hands, planning, 1:1, etc.)
- Date (from the transcript header or context)
Step 3: Extract long-term knowledge
Read the transcript and extract only durable knowledge - information that will still be relevant months from now. For each item, classify it:
What to extract
| Category | What to look for |
|---|
| Product direction | New product ideas, feature visions, strategic pivots, market positioning signals |
| Architecture/tooling | New tools, systems, or architectural patterns being built or adopted (how they work, why they were chosen) |
| Domain knowledge | Regulatory, scientific, or industry context that helps understand the business |
| Team/org changes | New hires, role changes, team restructuring, new responsibilities |
| Process changes | New workflows, changed policies, new vendor relationships that affect how work gets done |
| Client/market signals | Patterns in client needs or market dynamics (not individual deal details) |
What to ignore
| Category | Why |
|---|
| Status updates on in-flight work | Ephemeral - will be outdated in days |
| Deadlines and timelines | Ephemeral - stale immediately after the date passes |
| HR/benefits/admin announcements | Not relevant to project context |
| Small talk, introductions of existing team members | No knowledge value |
| Bug reports or incident details | Belong in issue trackers, not docs |
| Specific customer names and deal values | Business-sensitive, not engineering context (unless the client relationship reveals a product pattern) |
| Action items and to-dos | Belong in task trackers |
Step 4: Reconcile with existing docs
Read docs/index.md to understand existing documentation structure.
For each knowledge atom:
- Find the right file - check if an existing doc covers the topic
- If a doc exists - read it, update the relevant section. Merge new information naturally into existing prose. Do not create duplicate sections.
- If no doc exists - only create a new file if the knowledge does not fit into any existing doc.
- If content contradicts existing docs - update the doc with the newer information. Add the meeting date as source context (e.g., "as of April 2026 weekly sync").
Writing conventions
- Hyphens (-) instead of em-dashes in all text
- Full sentences with pronouns and normal punctuation
- No fluff - write for an AI assistant that needs facts, not prose
- Lead with the what, then the why
- Attribute ideas to people when it helps understand the organizational context (e.g., "Matt identified the need for..." or "Samantha proposed...")
Step 5: Update docs/index.md
If any new files were created, add them to docs/index.md under the appropriate section.
Step 6: Report
Output a concise summary of what was extracted and where it was written:
Meeting: [type] - [date]
Participants: [names]
Changes:
- Updated docs/<area>/engineering.md: [what was added]
- Updated docs/<area>/overview.md: [what was added]
- Created docs/<area>/new-topic.md: [why a new file was needed]
Skipped (ephemeral):
- [1-2 sentence summary of what was intentionally left out]
Key principles
- Err on the side of excluding. If you are unsure whether something is long-term relevant, leave it out. The user can always ask you to add more.
- Never add deadlines or dates as actionable items. Dates are only useful as temporal context for when a decision was made.
- Merge, don't duplicate. Always check existing docs first. The goal is to enrich existing documentation, not create parallel files.
- Preserve existing structure. When updating a file, respect its current organization. Add new sections where they naturally fit.
- Do not use em-dashes. Use hyphens, commas, or parentheses instead.