https://www.sebastianrotter.de/Practical articles about AI coding agents, context engineering, agent workflows and data engineering 2026-10-11T21:52:32+09:00 Sebastian Rotter https://www.sebastianrotter.de/ Jekyll © 2026 Sebastian Rotter /assets/img/favicons/favicon.ico /assets/img/favicons/favicon-96x96.png 2026-10-11T00:00:00+09:00 2026-10-11T00:00:00+09:00 https://www.sebastianrotter.de/posts/the-7-stage-skill-workflow-i-use-with-coding-agents/ Sebastian Rotter

In my previous article, I described the seven Markdown files I use to manage projects with coding agents. Those files keep the current status of the project, but they are not what moves it from idea to implemented code. For that, I use a sequence of skills I have set up in Codex: init-project grill-me or grill-with-docs create-system-design create-implementation-plan create-impleme...

2026-08-22T00:00:00+09:00 2026-09-05T20:22:31+09:00 https://www.sebastianrotter.de/posts/7-markdown-files-i-use-to-manage-projects-with-coding-agents/ Sebastian Rotter

When I started using AI coding agents, I managed most projects through conversation. That works well for tasks that fit into one session. It becomes harder when a project spans several sessions and a new agent needs to recover previous decisions, preferences, and unfinished work. I used to handle this with handover prompts. Those help, but important decisions still get buried, discovered tasks ...

2026-07-12T00:00:00+09:00 2026-09-05T20:22:08+09:00 https://www.sebastianrotter.de/posts/managing-agent-context-part-2-full-context-windows-compaction/ Sebastian Rotter

In part 1, I broke down what goes into an agent’s context window. This part looks at what happens when that window starts filling up, why compaction exists, and why I try not to rely on it too much. What happens when the context window fills up In practice, an agent with a full context window can look a bit like a person dealing with information overload: it loses its focus and mixes things u...

2026-06-27T00:00:00+09:00 2026-09-05T20:22:50+09:00 https://www.sebastianrotter.de/posts/the-next-stage-of-llm-ui-ux/ Sebastian Rotter

Anthropic announced Claude Tag on 23 June 2026, introducing it as a new way for teams to work with Claude. The product demos really impressed me because they show new ways for individuals as well as whole teams to work with Claude outside a terminal or designated chat interface. I believe they have given us a glimpse into the future where this way of working could be considered normal for both ...

2026-06-14T00:00:00+09:00 2026-09-05T20:21:41+09:00 https://www.sebastianrotter.de/posts/managing-agent-context-part-1-understanding-the-context-window/ Sebastian Rotter

One of the most important skills to develop if you want to use agents effectively is learning how to manage their context. This is also referred to as context engineering: deciding what information the agent can see and what should stay out. Before we can manage context well, it helps to understand what is actually inside an agent’s context window. Anthropic defines context as “the set of toke...