Your Agent's Rules Are Probably Wrong
Without a feedback loop, your agent skills teach stale patterns — or worse, propagate security rules that were never right.
4 posts
The plumbing that makes agent-assisted engineering actually reliable day to day: a 4-tier context architecture, catching drift before agent rule files quietly go stale, the decision tree for skills versus instructions versus scripts, and testing the infrastructure underneath it all with Terratest.
Without a feedback loop, your agent skills teach stale patterns — or worse, propagate security rules that were never right.
The architecture that cut per-session agent context by 85–89% — without losing a single useful rule.
Not all agent context is the same. The difference between skills, instructions, and scripts determines whether your agent stays focused.
Write automated tests against your infrastructure code using LocalStack to mimic the AWS cloud environment — no AWS account needed.