Exam Room · Advanced Generative AI Developer

Pop Quiz: Prompts Are Versioned Assets

· 4 min read

Exam-style

Thirty microservices each embed a copy of the same three Bedrock system prompts as string literals, so a single wording fix means thirty pull requests. The team wants one place to store prompts, version them, and have each service reference a specific version. What fits?

Reveal the answer

A. Bedrock Prompt management, with each service passing the prompt version ARN where a model ID normally goes

Prompt management stores a prompt as a managed resource carrying versions, variables, and the model and inference configuration to run it with. Converse and InvokeModel accept the ARN of a prompt version in place of a model ID, with promptVariables supplying the values, so a service names version 3 and holds no prompt text of its own. A wording fix becomes a new version, and each service moves to it on its own schedule. Parameter Store does retain the 100 most recent versions of a parameter, but a value caps at 4 KB on the standard tier and 8 KB on the advanced tier, and there is no variable substitution and no side-by-side variant testing. A Lambda layer has the same gap plus one more: a function names an exact layer version ARN, so new wording still means updating thirty function configurations, and layers apply only to .zip functions rather than container images. Knowledge Bases retrieve grounding documents for RAG rather than storing prompt templates. AgentCore Gateway turns APIs and Lambda functions into MCP tools, which is a tool registry, not a prompt registry.

Generative AI Development · part of The Exam Room

Q. Thirty services share prompts and you need versioning and reuse. What on Bedrock?

A. Bedrock Prompt management stores and versions prompts as managed resources. Converse takes the ARN of a prompt version where a model ID normally goes, so a service names a version instead of carrying the text. Prompt cachingReusing the model’s already-processed prefix (system instructions, fixed context) across calls so you don’t pay to re-read it every time. is a separate setting on the prompt, and it reduces response latency and input token cost when a long context repeats.

Why? Version prompts as managed resources, rather than copying strings across services.

These posts are LLM-aided. Backbone, original writing, and structure by Craig. Research and editing by Craig + LLM. Proof-reading by Craig.