Your next internal developer platform is a library of agent skills
What happens when AI agents become direct users of your infrastructure?
Michael Kutsch, Staff SRE and Team Lead for Cloud Foundations at PostHog, argues that the next internal developer platform may be a library of agent skills. Instead of forcing every task through a portal, his team is giving agents structured context, reusable workflows, and deterministic scripts they can call when reliability and governance matter.
The technical shift depends on a cultural one. Michael explains how PostHog's high-trust engineering culture lets teams contribute directly to infrastructure, why a giant instruction file produced inconsistent pull requests, and how a layered context system improved the results. He also shares how Grafana MCP helps teams investigate incidents and why infrastructure engineers are moving toward architecture and review.
In this episode:
- Why engineers at PostHog can act without layers of approval
- How product teams contribute directly to infrastructure
- Why agents need an index to the right context, not one giant prompt
- Where deterministic scripts belong inside agent workflows
- How agents can reduce the dashboard hopping of incident response
- Which parts of infrastructure engineering Michael expects to endure
Connect with Michael: https://www.linkedin.com/in/michael-kutsch-6445a483
Learn more about PostHog: https://posthog.com
#SRE #PlatformEngineering #AIEngineering #DevOps #ReliabilityEngineering