Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Triage: Is search returning results? Check Glean status page. Trigger: "glean incident runbook", "incident-runbook".
.claude/skills/jeremylongshore-glean-incident-runbook/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -8% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -3% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -17% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -18% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -5% | 0% |
Incident response procedures for Glean enterprise search integration failures. Covers search degradation, connector sync failures, indexing backlogs, and permission sync drift. Glean aggregates knowledge across all company tools, so incidents impact employee productivity across the entire organization. When search breaks or returns stale results, teams lose access to critical institutional knowledge. Classify severity immediately and follow the matching playbook below.
| Level | Definition | Response Time | Example | |-------|-----------|---------------|---------| | P1 - Critical | Search fully down or returning zero results | 15 min | All queries return empty, API 5xx errors | | P2 - High | Connector sync failed, content going stale | 30 min | Google Drive connector last synced 24h ago | | P3 - Medium | Indexing backlog or partial result degradation | 2 hours | New documents not appearing for 4+ hours | | P4 - Low | Permission sync drift or single datasource issue | 8 hours | One user sees docs they shouldn't access |
bash# Test search API health curl -s -o /dev/null -w "HTTP %{http_code}\n" \ -H "Authorization: Bearer $GLEAN_API_TOKEN" \ -H "Content-Type: application/json" \ -X POST https://your-domain.glean.com/api/v1/search \ -d '{"query": "test", "pageSize": 1}' # Check datasource connector status curl -s -H "Authorization: Bearer $GLEAN_API_TOKEN" \ https://your-domain.glean.com/api/v1/getdatasourceconfig \ -d '{"datasource": "DATASOURCE_NAME"}' | jq '.status' # Verify indexing queue depth curl -s -H "Authorization: Bearer $GLEAN_API_TOKEN" \ https://your-domain.glean.com/api/index/v1/getstatus | jq '.statistics'
echo $GLEAN_API_TOKEN | wc -cgetdatasourceconfig for each sourcemarkdown**Incident**: Glean Search [Outage/Degradation] **Status**: [Investigating/Identified/Mitigating/Resolved] **Started**: YYYY-MM-DD HH:MM UTC **Impact**: [Search unavailable / results stale since HH:MM / N datasources not syncing] **Current action**: [Connector re-auth in progress / Glean support engaged / manual re-crawl running] **Next update**: HH:MM UTC
| Incident Type | Detection | Resolution | |--------------|-----------|------------| | Search degradation | Empty results or low relevance scores | Check API health, verify index freshness | | Connector sync failure | Stale content, getdatasourceconfig shows error | Re-authorize connector, trigger manual crawl | | Indexing backlog | New docs not searchable after 4+ hours | Monitor queue depth, contact Glean if persistent | | Permission sync drift | Users see restricted docs or miss accessible ones | Audit datasource permissions, trigger permission re-sync |
See glean-observability for monitoring setup and connector health dashboards.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 23,423 | 14,737 | -37% | 1 | 1 | 0% | 3,408 | 3,141 | -8% | 0 | 0 | — |
case-02 | fail→pass | 21,043 | 12,198 | -42% | 1 | 1 | 0% | 2,735 | 2,650 | -3% | 0 | 0 | — |
case-03 | fail→pass | 21,924 | 15,768 | -28% | 1 | 1 | 0% | 4,007 | 3,331 | -17% | 0 | 0 | — |
case-04 | fail→fail | 24,987 | 31,531 | +26% | 1 | 1 | 0% | 4,403 | 5,495 | +25% | 0 | 0 | — |
case-05 | fail→fail | 32,240 | 18,663 | -42% | 1 | 1 | 0% | 3,675 | 4,048 | +10% | 0 | 0 | — |
case-06 | fail→fail | 20,166 | 24,603 | +22% | 1 | 1 | 0% | 3,506 | 5,525 | +58% | 0 | 0 | — |
case-07 | fail→pass | 13,036 | 9,516 | -27% | 1 | 1 | 0% | 2,717 | 2,230 | -18% | 0 | 0 | — |
case-08 | fail→pass | 15,334 | 3,640 | -76% | 1 | 1 | 0% | 2,056 | 1,958 | -5% | 0 | 0 | — |
case-09 | fail→pass | 16,524 | 13,072 | -21% | 1 | 1 | 0% | 2,253 | 2,989 | +33% | 0 | 0 | — |
case-10 | pass→pass | 10,793 | 1,635 | -85% | 1 | 1 | 0% | 1,091 | 1,483 | +36% | 0 | 0 | — |
case-11 | fail→pass | 13,981 | 14,118 | +1% | 1 | 1 | 0% | 2,555 | 2,962 | +16% | 0 | 0 | — |
case-12 | fail→pass | 10,042 | 4,312 | -57% | 1 | 1 | 0% | 1,611 | 1,481 | -8% | 0 | 0 | — |
case-13 | fail→pass | 13,343 | 4,175 | -69% | 1 | 1 | 0% | 1,324 | 1,828 | +38% | 0 | 0 | — |
case-14 | pass→pass | 10,945 | 7,613 | -30% | 1 | 1 | 0% | 1,724 | 1,560 | -10% | 0 | 0 | — |
case-15 | fail→pass | 18,710 | 6,752 | -64% | 1 | 1 | 0% | 2,182 | 2,126 | -3% | 0 | 0 | — |
case-16 | pass→pass | 13,085 | 7,162 | -45% | 1 | 1 | 0% | 1,275 | 1,460 | +15% | 0 | 0 | — |
case-17 | pass→pass | 14,268 | 9,272 | -35% | 1 | 1 | 0% | 2,220 | 2,010 | -9% | 0 | 0 | — |
case-18 | pass→pass | 7,703 | 7,227 | -6% | 1 | 1 | 0% | 1,325 | 1,501 | +13% | 0 | 0 | — |
case-19 | pass→pass | 14,723 | 4,328 | -71% | 1 | 1 | 0% | 1,438 | 1,910 | +33% | 0 | 0 | — |
case-20 | fail→pass | 11,419 | 8,241 | -28% | 1 | 1 | 0% | 1,802 | 1,675 | -7% | 0 | 0 | — |
case-21 | fail→pass | 15,298 | 2,471 | -84% | 1 | 1 | 0% | 1,805 | 1,524 | -16% | 0 | 0 | — |
case-22 | fail→pass | 11,498 | 1,988 | -83% | 1 | 1 | 0% | 1,938 | 1,484 | -23% | 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 +59 percentage points is the difference between those two pass rates over the 22 comparable cases.
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.