Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Build automated pipelines from Granola meeting notes to GitHub Issues, Linear tasks, Slack notifications, and documentation updates using Zapier and GitHub Actions. Trigger: "granola CI", "granola automation pipeline", "granola to github", "granola to linear", "meeting notes automation".
.claude/skills/jeremylongshore-granola-ci-integration/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 23% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 52% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 63% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 143% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 11% | 0% |
Build automated pipelines that process Granola meeting notes into development artifacts: GitHub Issues from action items, Linear tasks with team routing, Slack digests for stakeholders, and meeting logs in your repository. Uses Zapier as the middleware between Granola and dev tools.
yaml# Pipeline: Granola → Zapier → GitHub + Slack + Linear Trigger: App: Granola Event: Note Added to Granola Folder Folder: "Engineering" # Only process engineering meetings
Add a Code by Zapier step (JavaScript) to extract action items:
javascript// Zapier Code Step — Extract action items from Granola note const noteContent = inputData.note_content || ''; const meetingTitle = inputData.title || 'Untitled Meeting'; const meetingDate = inputData.calendar_event_datetime || new Date().toISOString(); // Extract action items: matches "- [ ] @person: task" or "- [ ] task" const actionRegex = /- \[ \] @?(\w+):?\s+(.+)/g; const actions = []; let match; while ((match = actionRegex.exec(noteContent)) !== null) { actions.push({ assignee: match[1], task: match[2].trim(), meeting: meetingTitle, date: meetingDate.split('T')[0], }); } // Extract decisions: lines starting with "- " under "## Decisions" or "## Key Decisions" const decisionSection = noteContent.match(/## (?:Key )?Decisions\n([\s\S]*?)(?=\n##|$)/); const decisions = decisionSection ? decisionSection[1].split('\n').filter(l => l.startsWith('- ')).map(l => l.replace('- ', '')) : []; output = [{ action_count: actions.length, actions: JSON.stringify(actions), decisions: decisions.join('; '), meeting_title: meetingTitle, meeting_date: meetingDate, }];
yaml# For each action item, create a GitHub issue Action: App: GitHub Event: Create Issue Repository: "your-org/your-repo" Title: "Meeting Action: {{task}} [{{date}}]" Body: | ## Context From meeting: **{{meeting}}** on {{date}} ## Task {{task}} ## Assigned To @{{assignee}} --- *Auto-created from Granola meeting notes* Labels: "meeting-action" Assignee: "{{assignee}}" # Must match GitHub username
Create a workflow triggered by Zapier via repository_dispatch:
yaml# .github/workflows/meeting-log.yml name: Update Meeting Log on: repository_dispatch: types: [granola-meeting] jobs: update-log: runs-on: ubuntu-latest steps: - uses: actions/checkout@v4 - name: Append to meeting log run: | MEETING_TITLE="${{ github.event.client_payload.title }}" MEETING_DATE="${{ github.event.client_payload.date }}" DECISIONS="${{ github.event.client_payload.decisions }}" ACTION_COUNT="${{ github.event.client_payload.action_count }}" mkdir -p docs/meetings cat >> docs/meetings/log.md << EOF ## ${MEETING_DATE} — ${MEETING_TITLE} - **Decisions:** ${DECISIONS} - **Action items created:** ${ACTION_COUNT} - **Source:** Granola AI EOF - name: Commit and push run: | git config user.name "Granola Bot" git config user.email "bot@granola.ai" git add docs/meetings/log.md git commit -m "docs: meeting log — ${MEETING_DATE}" || echo "No changes" git push
Trigger from Zapier using the Webhooks action:
yamlAction: App: Webhooks by Zapier Event: POST URL: https://api.github.com/repos/your-org/your-repo/dispatches Headers: Authorization: "Bearer {{github_pat}}" Accept: "application/vnd.github.v3+json" Body: event_type: "granola-meeting" client_payload: title: "{{meeting_title}}" date: "{{meeting_date}}" decisions: "{{decisions}}" action_count: "{{action_count}}"
yamlAction: App: Linear Event: Create Issue Team: Engineering Title: "{{task}}" Description: "From meeting: {{meeting}} ({{date}})\n\nAssigned: @{{assignee}}" Label: "meeting-action" Priority: "Medium"
yamlAction: App: Slack Event: Send Channel Message Channel: "#engineering-meetings" Message: | :memo: *Meeting Notes Ready:* {{meeting_title}} :calendar: {{meeting_date}} *Decisions:* {{decisions}} *Action Items Created:* {{action_count}} :point_right: Check Linear/GitHub for assigned tasks [View full notes in Granola]
Meeting ends → Granola enhances notes
→ Note added to "Engineering" folder
→ Zapier triggers
├→ Parse action items (Code step)
├→ Create GitHub Issues (per action item)
├→ Trigger GitHub Actions (update meeting log)
├→ Create Linear tasks (per action item)
└→ Post Slack summary (#engineering-meetings)| Error | Cause | Fix | |-------|-------|-----| | Zapier trigger not firing | Note not in the configured folder | Verify folder name matches exactly | | GitHub issue creation fails | PAT expired or insufficient scope | Regenerate PAT with repo scope | | Action items not parsed | Note format doesn't match regex | Adjust regex for your template's action item format | | Linear API error | Team name mismatch | Use Linear team ID instead of name | | Slack message empty | Note still processing | Add 2-minute delay as first Zap step |
Proceed to granola-deploy-integration for native app integration setup.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | pass→pass | 19,667 | 19,070 | -3% | 1 | 1 | 0% | 3,269 | 5,126 | +57% | 0 | 0 | — |
case-02 | fail→pass | 13,520 | 8,354 | -38% | 1 | 1 | 0% | 3,153 | 3,867 | +23% | 0 | 0 | — |
case-03 | fail→pass | 21,758 | 18,767 | -14% | 1 | 1 | 0% | 3,380 | 5,143 | +52% | 0 | 0 | — |
case-04 | pass→pass | 14,677 | 9,524 | -35% | 1 | 1 | 0% | 1,760 | 3,090 | +76% | 0 | 0 | — |
case-05 | fail→pass | 21,984 | 12,413 | -44% | 1 | 1 | 0% | 2,534 | 4,139 | +63% | 0 | 0 | — |
case-06 | pass→pass | 22,297 | 17,544 | -21% | 1 | 1 | 0% | 3,035 | 4,011 | +32% | 0 | 0 | — |
case-07 | pass→pass | 16,433 | 8,860 | -46% | 1 | 1 | 0% | 1,867 | 3,335 | +79% | 0 | 0 | — |
case-08 | fail→fail | 13,304 | 15,126 | +14% | 1 | 1 | 0% | 2,077 | 3,543 | +71% | 0 | 0 | — |
case-09 | fail→pass | 10,335 | 2,830 | -73% | 1 | 1 | 0% | 949 | 2,302 | +143% | 0 | 0 | — |
case-10 | fail→fail | 11,977 | 4,583 | -62% | 1 | 1 | 0% | 1,319 | 2,636 | +100% | 0 | 0 | — |
case-11 | fail→pass | 16,632 | 6,924 | -58% | 1 | 1 | 0% | 1,956 | 2,179 | +11% | 0 | 0 | — |
case-12 | fail→fail | 15,121 | 10,185 | -33% | 1 | 1 | 0% | 2,533 | 3,560 | +41% | 0 | 0 | — |
case-13 | fail→pass | 19,021 | 19,881 | +5% | 1 | 1 | 0% | 2,242 | 4,366 | +95% | 0 | 0 | — |
case-14 | fail→pass | 13,024 | 9,993 | -23% | 1 | 1 | 0% | 1,289 | 2,674 | +107% | 0 | 0 | — |
case-15 | pass→pass | 9,679 | 8,429 | -13% | 1 | 1 | 0% | 1,567 | 2,411 | +54% | 0 | 0 | — |
case-16 | fail→pass | 10,402 | 11,767 | +13% | 1 | 1 | 0% | 1,572 | 2,951 | +88% | 0 | 0 | — |
case-17 | pass→pass | 9,423 | 6,749 | -28% | 1 | 1 | 0% | 1,328 | 2,054 | +55% | 0 | 0 | — |
case-18 | pass→fail | 21,091 | 13,985 | -34% | 1 | 1 | 0% | 2,567 | 4,347 | +69% | 0 | 0 | — |
case-19 | fail→pass | 11,099 | 1,704 | -85% | 1 | 1 | 0% | 1,658 | 2,071 | +25% | 0 | 0 | — |
case-20 | fail→fail | 22,827 | 26,065 | +14% | 1 | 1 | 0% | 4,577 | 7,307 | +60% | 0 | 0 | — |
case-21 | fail→fail | 20,352 | 20,551 | +1% | 1 | 1 | 0% | 2,831 | 5,988 | +112% | 0 | 0 | — |
case-22 | fail→fail | 17,755 | 13,120 | -26% | 1 | 1 | 0% | 2,123 | 4,050 | +91% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +36 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.