Imported from jeremylongshore/tons-of-skills-marketplace (plugins/saas-packs/klaviyo-pack/skills/klaviyo-observability/SKILL.md). Install upstream with npx skills add jeremylongshore/tons-of-skills-marketplace --skill klaviyo-observability. Copyright stays with the author (MIT).

Klaviyo Observability

Overview

Comprehensive observability for Klaviyo integrations: Prometheus metrics for API call tracking, OpenTelemetry tracing, structured logging, and alerting rules tuned to Klaviyo's rate limits and error patterns. The pattern centers on one instrumentation wrapper that every Klaviyo call routes through, so metrics, traces, and logs stay consistent across profiles, events, and webhooks.

Prerequisites

• Prometheus or compatible metrics backend
• OpenTelemetry SDK installed (optional)
• Grafana or similar dashboarding tool (optional)
• klaviyo-api SDK installed

Key Metrics to Track

Instructions

Read any existing Klaviyo client code first, then build the layers in order. Each step writes one module; steps 5–6 wire the alerting and scrape endpoint.

1. Instrumented API wrapper — write src/klaviyo/instrumented-client.ts with the Prometheus counters, histogram, and gauge, exposed through a single instrumentedCall() helper.
2. Route every call through instrumentedCall(endpoint, method, () => ...) in the service layer so profile/event/webhook traffic is all counted.
3. OpenTelemetry tracing (optional) — add tracedKlaviyoCall() to emit spans with Klaviyo operation + error attributes.
4. Structured logging — add a pino logger with an email-redacting serializer.
5. Alert rules — drop prometheus/klaviyo-alerts.yml in place for error-rate, 429, latency, down, and low-headroom alerts.
6. Metrics endpoint — expose GET /metrics from the shared registry.

The wrapper is the load-bearing piece — the skeleton is:

export async function instrumentedCall<T>(
  endpoint: string,
  method: string,
  operation: () => Promise<T>
): Promise<T> {
  const timer = apiDuration.startTimer({ method, endpoint });
  try {
    const result = await operation();
    apiRequests.inc({ method, endpoint, status: 'success' });
    return result;
  } catch (error: any) {
    apiErrors.inc({ endpoint, status_code: error.status || 'unknown', error_code: error.body?.errors?.[0]?.code || 'unknown' });
    throw error;
  } finally {
    timer();
  }
}

Full source for all six steps — counters, tracing, logging, and the metrics endpoint — is in references/instrumentation.md. Alert rules and Grafana panels are in references/alerting.md.

Output

Applying this skill produces:

• src/klaviyo/instrumented-client.ts — Prometheus registry + instrumentedCall() wrapper
• src/klaviyo/tracing.ts — OpenTelemetry tracedKlaviyoCall() (optional)
• src/klaviyo/logger.ts — pino logger with PII-redacting serializers
• prometheus/klaviyo-alerts.yml — five alert rules (error rate, 429s, latency, down, low headroom)
• GET /metrics route exposing the registry in Prometheus text format

Once wired, curl localhost:PORT/metrics returns the klaviyo_* series, and the Grafana panels in references/alerting.md render request rate, error rate, P95 latency, and rate-limit headroom.

Error Handling

Examples

Instrument a profile upsert — wrap the SDK call so it counts toward klaviyo_api_requests_total and records latency:

const profile = await instrumentedCall('profiles', 'POST', () =>
  profilesApi.createOrUpdateProfile({
    data: { type: 'profile', attributes: { email: user.email, firstName: user.name } },
  })
);

Alert on rate-limit pressure — fire before you start getting 429s:

- alert: KlaviyoRateLimitLow
  expr: klaviyo_rate_limit_remaining < 20
  for: 30s
  labels: { severity: warning }
  annotations:
    summary: "Klaviyo rate limit headroom below 20 requests"

More worked examples — event tracking, tracing, structured logging, and the full alert group — are in references/instrumentation.md and references/alerting.md.

Resources

• references/instrumentation.md — full metrics, tracing, logging, and metrics-endpoint source
• references/alerting.md — Prometheus alert rules and Grafana dashboard panels
• Prometheus Best Practices
• OpenTelemetry Node.js
• pino Logger

Next Steps

For incident response, see the klaviyo-incident-runbook skill, which pairs these metrics and alerts with triage and escalation procedures.