PowerShell UserGroup InnSalzach Meeting – 13th August

Join us for an exciting PowerShell UserGroup InnSalzach session where we dive into the art of scripting!
Topic: The Agentic Operating Model: From Ad Hoc Agent Use to Versioned, Verifiable Work
Speaker: Raimund Andree
Date: 13th August
Many PowerShell practitioners have already used an Agent to refactor a script, create Pester tests, or investigate a failing build. The interesting question is no longer whether Agents can act. It is how to make their work repeatable across repositories and teams, verifiable before acceptance, and bounded when they touch real systems.

This session starts where introductory Agent Mode talks end. It introduces the Agentic Operating Model (AOM) as a practical discipline for moving from successful one-off interactions to versioned, reviewable work. We examine what the Agentic Loop — OBSERVE, PLAN, ACT, VERIFY, ITERATE — requires once an Agent can access an entire repository, invoke the terminal, and use external Tools.

Using PowerShell and GitHub Copilot Agent Mode in a live demonstration, we explore how Context engineering and a Memory Bank give the Agent durable project knowledge; how Instruction files, Custom agents, and Skills turn recurring expectations into reusable behavior; and how Pester, evals, and git diff provide different kinds of evidence.

We will also examine where Self-verification stops: tests can verify deterministic behavior, but they do not replace human review, architectural judgment, or containment of risky capabilities.

PowerShell is the demonstration medium, not the subject. The same model applies to infrastructure, runbooks, research, and correspondence. Attendees will leave with criteria for evaluating their current Agent practice and a concrete path from experimentation to disciplined daily use.

Learning Outcomes: After the session, participants will be able to:

- Assess their existing Agent use against the Agentic Operating Model.
- Explain how Context engineering, Git, and a Memory Bank affect Agent performance.
- Select appropriate uses for Instruction files, Custom agents, and Skills.
- Distinguish between deterministic tests, Agent evals, and human review.
- Identify where Agent capabilities require stronger human oversight and containment.