
Prompt Injection Through Tool Results
Text your tools return can carry instructions the model may follow. Why you can't filter that away, and how to design a C# MCP server so it can't do much harm.
NEW COURSE · ~10 H · .NET 10
Take the .NET app you already have and let AI use it safely: tools on your real services, every action audited, scoped by role, tested, traced and published.
Join the waitlist
Hear first when Build MCP Servers in C# opens, with early access and a launch discount.
One real .NET app, one MCP server

Take the .NET app you already have and let AI use it safely: tools on your real services, every action audited, scoped by role, tested, traced and published.
ABOUT
I've spent 12 years building .NET backends — most recently leading teams that design distributed systems on Azure. My day-to-day has been the unglamorous parts that keep production running: retries and circuit breakers, idempotent consumers, outbox patterns, tracing with OpenTelemetry, and decisions written down as ADRs.
That's the lens I bring to AI. An LLM call is just another unreliable dependency — it needs the same engineering discipline as everything else in your system. This course teaches exactly that, in C#.
Notes from production, not from the docs

Text your tools return can carry instructions the model may follow. Why you can't filter that away, and how to design a C# MCP server so it can't do much harm.

Large tool results cost tokens and make agents worse. Cursor paging, trimmed text, aggregate tools and a size budget test for C# MCP servers.

Write tools let an agent change real data. How to design them in C#: read-only by default, no bulk actions, idempotent writes and user confirmation with MRTR.