# Context Engineering Context engineering is the deliberate design of what information, instructions, tools, memory, metadata, and runtime state an AI system receives—and when it receives them. It extends prompt writing into an architectural concern: assembling the smallest high-value context that lets an agent act correctly. XDO's scoped variables, named targets, reusable capabilities, document structure, and natural-language-to-Do pipeline can be understood as an early deterministic context architecture. AutoSites likewise makes path, frontmatter, source, and collection metadata available as structured context. Related: [[wiki/Layered Context Architecture|Layered Context Architecture]], [[wiki/Just-in-Time Context|Just-in-Time Context]], [[wiki/Progressive Disclosure|Progressive Disclosure]], [[projects/XDO - Executable Markdown and Portable Workflows|XDO]]. Source: [Anthropic — Effective context engineering for AI agents](https://www.anthropic.com/engineering/effective-context-engineering-for-ai-agents)