Build Your Observability Skill
Your Task API runs in Kubernetes. Users report slow responses, but you have no visibility. Is it the database? The network? A memory leak? Without observability, you're debugging in the dark.
This chapter teaches you to instrument applications, build dashboards, and manage cloud costs. But you won't start by reading documentation and hoping you remember the right PromQL syntax. You'll start by owning a skill that encodes production patterns from Prometheus, OpenTelemetry, Jaeger, Loki, and OpenCost.
When you face a production incident at 2am, you won't search Stack Overflow. You'll invoke your skill, and it will generate the exact queries, dashboards, and alerts you need. That's the difference between learning observability and owning observability.
Step 1: Get the Skills Lab
Clone a fresh skills lab to ensure a clean starting point:
- Go to github.com/panaversity/claude-code-skills-lab
- Click the green Code button
- Select Download ZIP
- Extract the ZIP file
- Open the extracted folder in your terminal
cd claude-code-skills-lab
claude
Step 2: Write Your Learning Spec
Before building the skill, define what you want to accomplish. Create a file called LEARNING-SPEC.md:
# Observability Learning Spec
## What I Want to Learn
- Monitor Kubernetes applications with Prometheus metrics
- Trace requests across services with OpenTelemetry and Jaeger
- Aggregate logs with Loki and query with LogQL
- Define SLOs and create multi-burn-rate alerts
- Track cloud costs with OpenCost
## Success Criteria
- I can deploy a full observability stack in under 30 minutes
- I can write PromQL queries for the 4 golden signals
- I can instrument a FastAPI app and see traces in Jaeger
- I can identify the top 3 cost drivers in my cluster
This spec guides both you and Claude on what the skill should cover.
Step 3: Fetch Official Documentation
Use the /fetching-library-docs skill (or Context7 directly) to gather authoritative sources:
Using your fetching-library-docs skill, gather official documentation for:
1. Prometheus - metrics collection and PromQL
2. OpenTelemetry Python SDK - instrumentation
3. Grafana Loki - log aggregation and LogQL
I need patterns for Kubernetes monitoring, not just API references.
Claude fetches production-relevant patterns from official sources, not Stack Overflow answers from 2019.
Step 4: Create the Skill
Now build your observability skill with everything grounded in what you just fetched:
Using your skill creator skill, create a new skill for Kubernetes observability
and cost engineering. I will use it to monitor applications from basic metrics
to production SRE practices. Include:
- Prometheus installation and ServiceMonitor configuration
- PromQL patterns for 4 golden signals
- OpenTelemetry FastAPI instrumentation
- Loki LogQL query patterns
- SLO/error budget alerting
- OpenCost cost allocation
Use the documentation you just fetched - no self-assumed knowledge.
Claude will:
- Reference the official docs it gathered
- Ask clarifying questions (sampling rates, retention, alert thresholds)
- Create the complete skill with tested patterns
Your skill appears at .claude/skills/observability-cost-engineer/.
Step 5: Test Your Skill
Verify the skill works by asking it a question:
Using your observability-cost-engineer skill, write a PromQL query that shows
the P95 latency for the task-api service over the last hour, broken down by
endpoint.
If the skill returns a correct histogram_quantile query, it's working. If it hallucinates syntax, refine the skill with corrections.
What Happens Next
Each lesson in this chapter tests and improves your skill:
| Lesson | What You Learn | Skill Improvement |
|---|---|---|
| L01 | Three Pillars of Observability | Add decision framework: metrics vs traces vs logs |
| L02 | Prometheus + PromQL | Add recording rules and ServiceMonitor templates |
| L03 | Grafana Dashboards | Add dashboard JSON templates for golden signals |
| L04 | OpenTelemetry + Jaeger | Add FastAPI instrumentation code and sampling config |
| L05 | Loki + LogQL | Add structured logging patterns and log-trace correlation |
| L06 | SLIs, SLOs, Error Budgets | Add SLO calculation formulas and budget burn rates |
| L07 | Alerting | Add multi-burn-rate PrometheusRule templates |
| L08 | Cost Engineering | Add OpenCost queries and right-sizing recommendations |
| L09 | Dapr Integration | Add Dapr observability Configuration CRD |
| L10 | Capstone | Complete integration testing and skill finalization |
By chapter end, your skill contains production-tested patterns for the entire observability stack.
Try With AI
Prompt 1: Explore the Three Pillars
I'm about to learn Kubernetes observability. Before diving into tools, help me
understand the landscape. Ask me about a recent debugging experience where I
wished I had more visibility. Based on that, explain which of the three pillars
(metrics, traces, logs) would have helped most and why.
What you're learning: Self-assessment of current observability gaps through Socratic dialogue. Your skill will encode these patterns, but understanding the "why" helps you know when to apply each pillar.
Prompt 2: Validate Your Skill's Coverage
I just created an observability skill. Here's what it covers: [paste your skill's
key sections]. Review this against production SRE requirements. What's missing?
What would a senior SRE add? Ask me about my deployment environment so your
recommendations are specific to my needs.
What you're learning: Skill validation through expert review. You're teaching your AI partner about your context so it can identify gaps specific to your situation.
Prompt 3: Plan Your Observability Stack
I'm deploying Task API to Kubernetes and need to choose observability tools.
I know about Prometheus, Grafana, Jaeger, and Loki from my new skill. Help me
prioritize: If I can only deploy two tools this week, which two give me the
most value? Ask me about my team size, budget, and biggest operational pain
points.
What you're learning: Strategic prioritization under constraints. Not every stack needs every tool. Your skill contains patterns for all; this exercise teaches you which patterns apply to your situation first.
Safety Note
Observability tools collect sensitive data about your applications and infrastructure. When configuring Prometheus scrape targets, Loki log streams, or OpenTelemetry traces, ensure you're not capturing secrets, PII, or credentials. The patterns in your skill include security considerations, but always validate against your organization's data handling policies.