Context Engineering

The architectural discipline of designing and maintaining persistent file-based context — rather than ephemeral session prompts — for AI agents. Where prompt engineering is a per-session communication skill, context engineering treats project specifications, coding conventions, test policies, and domain knowledge as durable structured artifacts (e.g., CLAUDE.md, cursor-rules.md) that give any agent consistent, high-quality context across sessions.

Emerging as a core competency in AI-augmented software development in 2026, context engineering shifts the developer’s role from “instruction giver” to “context architect,” with memory file design becoming a distinct engineering discipline in agentic workflows.