Tag: Architecture

The same governance patterns, unmodified, on LangGraph and Microsoft Agent Framework — with parity tests and the footguns found in each.

Individual patterns are easy to believe in. Running twelve together on one agent found a real gap none of their own tests could have caught.

A spreadsheet of agents drifts the day after you write it. Make the profile enforced: owners, declared tools, expiry, and privilege-drift detection.

An unbounded tool response costs 17x more tokens for the same answer. A runnable auditor for the four tool-design mistakes that make agents guess.

Self-reported completion is not evidence. Independent deterministic verifiers, feedback re-injection, and why an LLM judging itself is theatre.

At-least-once execution is the default everywhere. Checkpoints make resumption cheap; idempotency keys make it safe. Runnable code and the bug I hit.