Encoding runbooks and golden-path standards as agent skills, with humans at every irreversible step
A platform team's hardest-won knowledge, such as which telemetry to check first, which cluster quirks to rule out and what a correct resource definition looks like, usually lives in a few people's heads. I've been working with AI agent skills that encode that knowledge. Investigations then follow the same evidence trail every time, and provisioning requests become pull requests that already meet platform standards.
| Skill | What it does |
|---|---|
| Troubleshooting | Classifies the issue, gathers evidence across APM, Kubernetes, Terraform state, cloud and pipelines, then correlates it and drafts findings |
| Provisioning | Writes Terraform, Helm values, pipeline YAML and catalog entries that follow golden-path conventions |
| Containerise an existing app | Reads an app's source and proposes a Dockerfile and Helm values |
| Secret-expiry triage | Cross-checks identity, telemetry and pipeline usage to recommend rotate, delete or suppress |
| Documentation | Writes READMEs and wiki pages to the team's documentation standard |
The goal isn't replacing platform engineers. It's making the team's best practice the default for everyone, including the agent.
About this case study: this write-up is drawn from my hands-on platform engineering experience. Organisation names, system names, figures and timelines have been changed or generalised, and all code, configuration and queries are illustrative examples written for this page.