Type: url
Original: https://martinfowler.com/articles/exploring-gen-ai/context-engineering-coding-agents.html
Status: generated
Summary
In this February 2026 “Exploring Gen AI” memo, Birgitta Böckeler gives a primer on the context configuration features of coding agents, separating reusable prompts (instructions and guidance) from what she calls context interfaces (tools, MCP servers, skills), and classifying features by who decides to load them: the LLM, a human, or the agent software. Using Claude Code’s features as of January 2026 as the worked example (CLAUDE.md, rules, slash commands, skills, subagents, MCP servers, hooks, plugins), she recommends keeping context small and building it up gradually, lists the difficulties of sharing context configurations, and warns that context engineering offers only an illusion of control because outcomes remain probabilistic.