From first-line IT to systems that can act.

Learn how computers run services, how teams ship changes, and how to operate systems that can act. Start with the basics or jump straight to the sharp edge.

Choose your route Read the field notes

Choose your route

Start where you are.

No assumed background. Learn the building blocks first, or skip to the operational problems that matter when you already run systems.

Start with the job

Useful on a bad day

Five entry points for the moments where an automated system needs evidence, boundaries or a safe way back.

Five task-first starting points. Browse the full index for Kubernetes, GitOps, IaC and classic SRE material.

Explore the full library

Field notes

Read this before you automate

Short primary-source reading for building systems that can change things without becoming impossible to inspect or recover.

  • Start here
    GitHub Agentic Workflows

    Read the permissions, safe-output and lock-file model before letting a coding agent investigate CI or open changes in a repository.

    docs.github.com · GitHub documentation

    Guide Primary
  • Start here
    GenAI semantic conventions

    A shared naming and event model for model calls and agent workflows. Agree the telemetry shape before building dashboards or alerts.

    github.com · OpenTelemetry

    Standard Primary
  • Specification
    SLSA provenance

    A framework for increasing supply-chain guarantees. Start with provenance and verification when an agent or pipeline can change the artifact.

    slsa.dev · SLSA specification

    Spec Primary
  • Tooling
    Sign and verify release artifacts

    Sigstore's documentation for signing software and verifying who or what produced it, without turning key management into a second project.

    docs.sigstore.dev · Sigstore documentation

    Tooling Primary
  • Policy
    Policy as a separate decision

    OPA separates policy decisions from the services that enforce them. That boundary matters when an agent can propose or execute a change.

    openpolicyagent.org · OPA documentation

    Policy Primary
  • Classic
    Monitor the service, not the vanity metric

    The SRE chapter on latency, traffic, errors and saturation, and why symptoms are usually better alert targets than guessed causes.

    sre.google · Google SRE Book

    Book Primary
  • Failure
    Check the repository before the agent does

    OpenSSF Scorecard turns common supply-chain controls into repeatable checks. Use the results to decide what an agent may safely touch.

    scorecard.dev · OpenSSF Scorecard

    Security Primary
  • Practice
    Treat tool protocols as production interfaces

    The MCP specification defines the surface between a model and tools. Read the transport, lifecycle and security sections before wiring in real permissions.

    modelcontextprotocol.io · MCP specification

    Spec Primary

The operator's loop

Observe, constrain, recover

A small loop for systems where both code and agents can make changes.

Editorial policy

Why these references?

A few rules keep this from becoming another link dump.

An agent adds a decision-maker to the delivery loop. That makes observability, permissions, provenance and recovery more important than another list of automation tools.

The index is static and hand-curated. It prefers upstream documentation, specifications and durable references, and intentionally avoids publishing fake update dates.

Pick the loop that matches the problem: Observe when you lack evidence, Constrain when automation can cause harm, or Recover when the failure path is unclear.

Some fundamentals outlast their tooling. A reference stays when it explains failure, trade-offs or an operating practice better than a newer product page.

No. It is a map for choosing the next useful source. Follow one path until you can apply it to a real service, then come back for the next gap.

There is no submission workflow yet. The list stays deliberately small until there is a useful way to review suggestions and keep links maintained.