# Context-as-Code Context-as-Code is the practice of treating AI context definitions as version-controlled, reviewable, testable code rather than ephemeral prompt text. Just as Infrastructure-as-Code brought software engineering discipline to infrastructure, Context-as-Code brings the same discipline to AI context management. ## Manifestations - **CLAUDE.md / AGENTS.md files**: AI behavior rules, conventions, and context stored in version-controlled files alongside code - **Skill definitions**: codified procedures (SKILL.md files) that define exactly how AI should perform specific tasks - **Agent configurations**: agent identity, memory, and capabilities defined in structured files - **MCP server configs**: tool and integration definitions stored as configuration - **Hook definitions**: event-driven behaviors defined in settings files ## Why it matters When context is code: - It can be **reviewed** in pull requests (someone changing AI behavior should be reviewed like any code change) - It can be **versioned** (you can roll back to a previous AI configuration) - It can be **tested** (verify that context changes produce expected outputs) - It can be **shared** (teams inherit and extend context through standard version control workflows) - It can be **audited** (compliance teams can inspect exactly what context AI operates with) ## Connection to the context management hierarchy Context-as-Code is the implementation pattern that makes [[Team Context Management (TCM)]] and [[Enterprise Context Management (ECM)]] practical. Without it, context management stays informal and fragile. With it, context management becomes an engineering discipline. At the [[Personal Context Management (PCM)|PCM]] level, a vault like this one (with CLAUDE.md, skills, agent configurations, and memory systems) is already practicing Context-as-Code. ## References - ## Related - [[Context Engineering]] - [[Personal Context Management (PCM)]] - [[Team Context Management (TCM)]] - [[Enterprise Context Management (ECM)]] - [[Harness Engineering]] - [[Claude Code]] - [[AI Agent Skills]] - [[AI Agent Identity]] - [[Model Context Protocol (MCP)]]