Maction

Step 1: Parse input

• Customer name (required): fuzzy-match against templates/customers.yaml names
• Meeting notes: the rest of the input. Can be:
  • Pasted text (Granola summary, transcript, bullet points)
  • File path to a .md or .txt file
  • Slack thread URL (e.g., https://coreweave.slack.com/archives/C06NX624Q9G/p1234567890) — fetch the full thread using the Slack skill, then process the thread content as meeting notes
  • If notes are very short (< 50 words), ask user if they want to paste more

Slack URL handling: If the input contains a Slack thread URL, fetch the thread first:

uv run --project .claude/skills/slack python .claude/skills/slack/scripts/threads.py --url "<slack_url>" --pretty

Use the thread content as the meeting notes input for all subsequent steps. Include the Slack URL as a source link in the Confluence meeting notes entry.

Step 2: Extract action items

Analyze the meeting notes and extract action items. For each action item, determine:

Step 3: Extract RAID items

Analyze the meeting notes for RAID signals:

• Risks: mention of churn, competitor, champion leaving, declining usage, renewal concern, "worried about", "risk of"
• Assumptions: implicit assumptions ("assuming they'll renew", "expecting migration by Q3", "plan is to...")
• Dependencies: "blocked on", "waiting for eng to", "need product to ship", "depends on"
• Issues: typically already tracked in Jira -- only flag explicit new issues not yet in Jira

For each RAID item, determine:

Step 3.5: Extract customer feedback signals

Analyze the meeting notes for customer-originating feedback -- their words, their sentiment, their satisfaction signals. This is NOT your internal assessment of the customer; it's what THEY expressed.

For each feedback item, determine:

If no feedback signals are detected, skip this step silently -- not every meeting has customer feedback.

Step 3.6: Compose meeting notes entry

Prepare a Confluence meeting notes entry in the standard FE format. The entry is a collapsible section with:

• Title: YYYY-MM-DD -- <Short Title> (<Participants>)
• Source Links: Links to Gong call, Granola notes, Slack threads (from the input)
• Summary: 3-5 bullet points covering what was discussed
• Action Items: List of actions extracted in Step 2 (with owners)
• Key Decisions: Any decisions made during the meeting
• RAID Signals: Any risks/dependencies flagged in Step 3

The Confluence storage format for one entry:

<ac:structured-macro ac:name="expand" ac:schema-version="1">
  <ac:parameter ac:name="title">YYYY-MM-DD -- Short Title (Participants)</ac:parameter>
  <ac:rich-text-body>
    <p><time datetime="YYYY-MM-DD" /></p>
    <p><strong>Source Links:</strong></p>
    <ul>
      <li><p>Gong: <a href="URL">Call title</a> (or "Not recorded")</p></li>
      <li><p>Granola: <a href="URL">Notes title</a> (or "None")</p></li>
      <li><p>Slack: <a href="URL">Thread context</a> (or "None")</p></li>
    </ul>
    <h3>Summary</h3>
    <ul><li><p>Key point 1</p></li>...</ul>
    <h3>Action Items</h3>
    <ul><li><p>[Owner] Action description -- Due: YYYY-MM-DD</p></li>...</ul>
    <h3>Decisions</h3>
    <ul><li><p>Decision made</p></li>...</ul>
  </ac:rich-text-body>
</ac:structured-macro>

Step 4: Present proposals to user

Show all extracted items in a clear format before creation:

Actions from [Customer] meeting ([date]):

PROPOSED ACTIONS (for [Customer] Actions project):
1. [P1] Send SDK migration guide to customer -- Due: 2026-04-01 -- Owner: me -- Section: To Do
2. [P2] Follow up on WB-1234 fix status (WB-1234) -- Due: 2026-03-28 -- Owner: me -- Section: Waiting on Eng
3. [P3] Customer to provide access credentials -- Due: 2026-04-05 -- Owner: customer -- Section: Waiting on Customer

PROPOSED RAID ITEMS (for [Customer] RAID Log):
R1. [Risk] Champion VP Eng may be leaving -- Impact: High, Likelihood: Medium -- Source: Cadence call 2026-03-24
D1. [Dependency] Blocked on eng shipping SDK v2.1 hotfix -- Impact: High, Likelihood: High -- Source: Cadence call 2026-03-24

CONFLUENCE MEETING NOTES (will be appended to [Customer] - Meeting Notes):
Title: 2026-03-24 -- Cadence Call (Allan, Customer Lead)
Summary: 3 bullet points
Sources: Granola link, Gong link

[If feedback found:]
CUSTOMER FEEDBACK (will be appended to [Customer] - Feedback):
F1. [Complaint] "We've been waiting three weeks for this fix" -- Status: OPEN

Confirm: create all (y), select items (s), or cancel (n)?

Step 5: Create confirmed items

For user-confirmed action items:

uv run --project .claude/skills/asana python .claude/skills/asana/scripts/mutate.py create \
  --project-gid <action_tracker_id> --name "<task name>" \
  --section "<section>" --assignee me --due <YYYY-MM-DD> --priority <P0-P3> \
  --notes "Source: Meeting notes [date]\n\nContext: <relevant excerpt from notes>" \
  --pretty

For user-confirmed RAID items:

uv run --project .claude/skills/asana python .claude/skills/asana/scripts/mutate.py create \
  --project-gid <raid_tracker_id> --name "<description>" \
  --section "<Risks|Assumptions|Issues|Dependencies>" --assignee me \
  --notes "Source: <source>\n\nContext: <relevant excerpt from notes>" \
  --pretty

After creation, set custom fields on RAID items (Category, Impact, Likelihood, Status=Open, Source, Visibility) via mutate.py update with --custom-fields or through the Asana API directly.

Step 6: Publish meeting notes to Confluence

Look up confluence_pages.meeting_notes from the customer's entry in templates/customers.yaml. If configured:

1. Fetch the current page to get the existing body:

   uv run --project .claude/skills/confluence python .claude/skills/confluence/scripts/pages.py get --id <meeting_notes_page_id> --pretty
   

2. Prepend the new collapsible entry (from Step 3.6) to the TOP of the existing body, after any intro text. New entries go at the top so the most recent meeting is first.

3. Update the page with the combined body:

   uv run --project .claude/skills/confluence python .claude/skills/confluence/scripts/pages.py update \
     --id <meeting_notes_page_id> --title "<existing title>" \
     --body "<intro text><new entry><existing entries>" --pretty
   

If confluence_pages is not configured for this customer: Skip silently and note in the summary that Confluence publishing was skipped.

Step 6.5: Publish feedback to Confluence

Only run this step if feedback items were extracted in Step 3.5.

Look up confluence_pages.feedback from templates/customers.yaml. If configured:

1. Fetch the current page to get the existing body.

2. For each feedback item, create an expandable entry in the same format as the FE team's existing pages:

   <ac:structured-macro ac:name="expand" ac:schema-version="1">
     <ac:parameter ac:name="title">YYYY-MM-DD [Type] Short description [Updated YYYY-MM-DD]</ac:parameter>
     <ac:rich-text-body>
       <p><time datetime="YYYY-MM-DD" /></p>
       <p>Summary: <em>"Direct customer quote if available"</em></p>
       <p>Context from meeting notes.</p>
       <p><strong>Links:</strong></p>
       <ul>
         <li><p>Gong: <a href="URL">Call</a> (or None)</p></li>
         <li><p>Granola: <a href="URL">Notes</a> (or None)</p></li>
       </ul>
       <p><strong>Status:</strong></p>
       <ul><li><p>Current status and any planned response.</p></li></ul>
     </ac:rich-text-body>
   </ac:structured-macro>
   

3. Insert the new entry under the appropriate section header (OPEN or BACKLOG) in the existing page body.

4. Update the page.

If no feedback signals found: Skip this step entirely. If confluence_pages is not configured: Skip silently.

Step 7: Summary

After all items are created, output a summary:

Created:
- 3 action items in [Customer] Actions project
- 2 RAID items in [Customer] RAID Log
- Meeting notes published to Confluence (2026-03-24 -- Cadence Call)
- 1 feedback item published to Confluence
Links: [task URLs, Confluence page URLs]

Granola Format Notes

Granola summaries typically have these sections:

• Summary / Key Points
• Action Items (often pre-extracted but may be incomplete)
• Decisions Made
• Discussion Topics

When processing Granola output, use the pre-extracted action items as a starting point but also scan the full summary for missed actions and RAID signals. Granola often misses RAID-type items since it focuses on explicit action items.

RAID Detection Heuristics

Context matters: "risk mitigation" is different from "there's a risk of churn". Use surrounding sentences to disambiguate.

Edge Cases

• RAID project not configured (raid_tracker_id is PLACEHOLDER): extract RAID items and display them but warn that they can't be created in Asana. Suggest running /asana setup-raid-project first.
• Customer Actions project not configured (action_tracker_id is PLACEHOLDER): same warning, suggest /asana setup-project.
• Meeting notes are very short (< 50 words): ask user if they have more context to share before extracting.
• No action items or RAID items found: "No actionable items or RAID signals detected in these notes. If you expected items, try providing more context."
• Ambiguous customer: if customer name matches multiple entries, ask user to clarify.

Safety Rules

• User confirms ALL items before creation. Never auto-create.
• Show exactly what will be created (name, project, section, fields) before confirmation.
• Never modify existing tasks -- only create new ones.
• RAID items default to Internal visibility.

Troubleshooting

Related Skills

• /asana -- base skill for task creation
• /raid -- RAID items created here appear in /raid view
• /ghosted -- tasks created in "Waiting on Customer" are tracked by /ghosted
• /cadence-prep -- actions and RAID items created here feed into cadence-prep agendas