Christopher Harrison

On his personal site, Harrison describes spending most of his career as a trainer, presenter or content creator. He says his teaching starts with deconstructing the parts he found difficult and showing how they became understandable. His technical interests include web development, TypeScript, Python and Django, with work across chatbots, DevOps and machine learning. GitHub’s author archive shows that approach applied to command-line agents, browser-assisted debugging, legacy-code explanations and concurrent agent sessions.[5][1]

Teaching the context around a coding task

Harrison’s guide to customizing Copilot separates four useful kinds of context. Repository instructions describe standing conventions, skills package repeatable procedures, custom agents supply focused guidance and Model Context Protocol servers connect outside tools. The distinction helps a developer decide where a piece of knowledge belongs instead of adding every instruction to each prompt. The article uses a concrete starting point: capture one project convention in a repository instruction file, then observe how it affects the next task.[3]

Maintenance as an inspectable workflow

His Dependabot tutorial uses dependency-update pull requests as a recurring task for an agent. The workflow gathers open updates, groups them by risk and produces a summary so a developer can decide what needs attention. A larger upgrade can then continue in a session carrying the existing context. Saved run histories make it possible to inspect what happened on previous attempts. The lesson is about teaching an automation a bounded maintenance procedure while keeping the resulting decisions visible.[4]