How I turned repository context into a PR Check
I started noticing a pattern while working with AI coding agents. The agent could write code surprisingly quickly, but the quality of the change depended heavily on how well it understood the repository. That led me to look at something that had existed in our repositories long before AI coding agents became common: engineering context. Things like: why a service is structured in a particular way which layer should own a database operation how services communicate which queues or events are involved how to run a particular test which constraints a developer needs to know before changing something what authentication method should be strictly followed while exposing APIs Some of this knowledge was in documentation | Some was in tests | Some was in architecture decisions done when developers were not in the meeting | lot of it was simply in people’s heads | and at last only one of the technical expert knew all the gotchas….. on how something works end to end :) ...