11 Context engineering for software development

This chapter provides a practical view of context engineering in one of its most important application domains: software development. Modern AI support for software work has evolved from autocomplete and conversational copilots to AI-native IDEs, coding agents, and agentic development systems. At the same time, AI is no longer limited to writing code. It can support many activities across the software development lifecycle (SDLC), including requirements analysis, architecture, implementation, testing, documentation, maintenance, and operations.

11.1 AI-assisted coding

11.1.1 Stage 0: traditional development

11.1.2 Stage 1: autocomplete

11.1.3 Stage 2: conversational copilots

11.1.4 Stage 3: AI-native IDEs

11.1.5 Stage 4: coding agents

11.1.6 Stage 5: agentic systems

11.2 AI-driven software development lifecycle

11.3 Core context engineering applied to software development

11.3.1 Agent skills for the SDLC

11.3.2 Instruction artifacts in coding agents

11.3.3 External knowledge for software development

11.3.4 MCP servers for the SDLC

11.3.5 Memory for coding agents

11.3.6 Reusable prompts as context assets

11.3.7 Specification-driven development

11.3.8 Context management for coding agents

11.3.9 Context orchestration for coding agents

11.4 Harness for software development

11.4.1 Evaluation

11.4.2 Observability

11.4.3 Governance

11.4.4 Operations

11.5 Hands-on

11.5.1 SDD with Spec Kit

11.6 Summary