Version control tracks changes to prompts and other AI context assets so teams can review updates, trace deployed configurations, and restore earlier versions. Store prompt templates separately from application code to inspect diffs and deploy changes independently. section 6.2.2 section 9.3.7
Apply software engineering controls to context changes:
- Keep context assets in a central Git repository with an auditable history and peer review. section 9.3.7
- Run regression evaluations before merging or deploying prompt and retrieval changes. section 6.2.2 section 9.3.7
- Promote validated configurations across development, staging, and production, with rollback available. section 9.3.7
Listing 9.1 Example directory structure for context-as-code artifacts
context/
prompts/
retrievers/
policies/
tools/
schemas/
memory/
evals/
environments/
dev/
test/
prod/
A representative repository layout separates prompts, retrieval, policies, tools, schemas, memory, evaluations, and environment-specific configurations. section 9.3.7



































