Summary
You now have the building blocks.
- Agents — A simple loop driven by a well-crafted system prompt.
- Tools — File access, command execution, and MCP servers give the agent the ability to act — and secrets connect it to your systems.
- Bots — Make your agent available where you already communicate, and let agents collaborate with each other.
- Skills — On-demand instructions make your agent sharper and more capable.
- Files — Connect your agent to the right files and data from multiple sources.
- Memory — File-based notes and a curated memory index let your agent remember across conversations and learn over time.
- Proactive Agents — Schedules and heartbeats turn your agent from a tool you use into a teammate that works alongside you.
With these concepts, you can design a powerful team of agents — each with a clear role, the right tools, and the ability to act on your behalf.
Next: the craft of designing agents
Section titled “Next: the craft of designing agents”Now that you understand the building blocks, the Agent Design Guide teaches you how to use them well. It covers the craft of writing good instructions — how to structure a system prompt, what to put in skills versus instructions, how to design schedules and heartbeats, and the common patterns and anti-patterns that separate a well-designed agent from a mediocre one.
Practice and iteration
Section titled “Practice and iteration”Building great agents seldom happens on the first attempt. The system prompt needs tuning. The skills need refining. The heartbeat prompts need adjusting as you discover how the agent behaves in different situations.
This is normal — and it is the core of agent engineering. Each iteration makes your agents better. The concepts in this guide give you a clear framework to work within, but the real learning comes from building, testing, and improving.
Get started
Section titled “Get started”Pick an agent platform and start building. Start simple — one agent, one clear responsibility, a few tools. Then expand from there as you learn what works.
If you are looking for a platform built from the ground up around these exact concepts, consider CloudNova. It is designed to make agent engineering accessible and powerful — so you can focus on designing great agents, not on the technical infrastructure underneath.
Whatever you choose, the most important step is the first one. Start building.