You can and should reuse shared context elements — your audience profile, your brand voice, your program overview — across multiple agents. But each agent still needs its own task-specific layer that defines its unique role, job scope, and limits. Shared context is the foundation; task-specific instructions are the differentiator.
Think of Context in Layers
An efficient multi-agent setup treats context like a modular system. Some elements are universal — they apply to every agent that represents your program. Other elements are role-specific — they apply only to the agent performing a particular task. When you structure context this way, you write the universal elements once and reuse them. You write the role-specific elements once per agent. Updates to the universal layer flow through to all agents automatically.
Think of it like a franchise. Every location uses the same brand guidelines, the same core menu, and the same customer service principles. But the weekend manager and the morning barista have different training for their specific roles. Shared foundation, differentiated function.
What to Share Across Agents
Your audience profile — who your students are, their experience level, their goals — is worth including in every agent context because it shapes how every agent communicates. Your brand voice — warm, direct, jargon-free — applies to every agent that interacts with your students. Your program overview — the high-level structure of what you offer — gives every agent enough background to orient students correctly even if the agent’s primary job is something else entirely.
Create a “shared context block” as a saved document. Every new agent gets this block pasted into its system prompt before the role-specific instructions begin. When your program changes — new cohort, new pricing, new module names — you update the shared block once and carry it forward to all agents. This is dramatically more maintainable than individually updating six different system prompts every time something changes.
What Must Stay Role-Specific
Job scope, constraints, and escalation rules must be written specifically for each agent’s role. Your content creation agent and your student support agent share the same audience and brand voice, but they have completely different jobs, different limits, and different escalation triggers. Mixing those in a generic shared prompt produces agents that are confused about their scope and wander into territory they were not designed for.
The Simple Rule
Build a shared context block covering audience, voice, and program overview. Use it as the foundation for every agent. Write a role-specific layer for each agent on top of that foundation. Update the shared block when your program changes. Update the role-specific layer when the agent’s job changes. One block shared, one layer unique — that architecture scales cleanly as your agent team grows.