Beginner7 min

System Prompt

Shape how your agent approaches tasks with a system prompt. Static strings for fixed personas, SystemMessage for provider features like prompt caching, and @dynamic_prompt for runtime-generated prompts.

Quick Reference

→system_prompt='...' sets a static instruction for every invocation
→Pass SystemMessage instead of str for provider-specific features (e.g. Anthropic cache_control)
→@dynamic_prompt middleware generates the prompt at runtime from state or context
→No system_prompt = agent infers task from messages only
→Dynamic prompts via context_schema: pass user_role or other metadata at invoke time

Static System Prompt

The simplest form: pass a string to system_prompt and it becomes the first message in every conversation. Use it to set the agent's persona, define its scope, and establish behavioral constraints.

String system prompt

SystemMessage with cache_control

When your system prompt is large — a full document, a long ruleset, a book — pass a SystemMessage instead of a string. This unlocks provider-specific features like Anthropic's prompt caching, which caches the content block and reuses it across requests, reducing both latency and cost.

@dynamic_prompt Middleware

For prompts that depend on runtime context — user role, language preference, current task — use the @dynamic_prompt decorator. It runs before each model call and returns a string that becomes the system prompt for that specific invocation.

Sign in to read this article

This is a premium article. Sign in with your Google account to continue.