SRE Engineer gives the agent a reliability workflow: assess the current state, define SLIs and SLOs, confirm they match user expectations, build monitoring and alerting, automate repetitive work and test resilience with chaos experiments.
The skill includes an error budget calculation, a multiwindow burn-rate alert rule for Prometheus, PromQL queries for the four golden signals and a Python script that restarts a deployment when its error rate passes a threshold.
What You Get
- A six-step reliability workflow with a user-expectation check on SLO targets
- A worked error budget calculation for a 99.9 percent availability SLO
- A Prometheus multiwindow burn-rate alert rule with fast and slow burn tiers
- PromQL queries for latency, traffic, errors and saturation
- A Python toil-automation example that checks a Prometheus error rate and restarts a deployment
- Reference files for SLO and SLI management, error budget policy, monitoring, automation and incidents
Common Use Cases
- Choosing SLIs and SLO targets for a service
- Writing burn-rate alerts with runbook links
- Reducing on-call toil with automation scripts
- Running blameless postmortems and chaos experiments
What Makes It Unique
It joins the policy side of SRE (error budgets, blameless postmortems, a limit on toil) with working configuration: alert rules, queries and a remediation script. Its constraints reject SLOs without user impact justification and alerts without runbooks.
Under The Hood
The skill is a SKILL.md with five reference files that the agent loads by topic. It ships no scripts of its own, but the Python example inside it queries a Prometheus HTTP API and runs kubectl rollout restart on a deployment. If you adopt that script, it reaches your Prometheus server and changes your cluster. The instructions also call for running chaos experiments, which deliberately cause failures in a running system and need your approval.