In depth
The system prompt is the contract between you and the LLM. It says what the agent is, what it should do, what it should not do, and how it should format its output. Most of the engineering work in agent design happens here, not in framework code. Writing a focused system prompt is more important than picking a more powerful model. On Digitorn the system prompt lives in YAML as a multi-line string, which means you iterate on it with hot reload instead of process restarts.
What it looks like in practice
YAML
1system_prompt: |2 You are a research agent. For each question:3 1. Search the web for 5 sources.4 2. Cross-reference and write a 600-word synthesis.5 3. Cite each claim with [domain.com](url).6 Be terse. No filler.Related concepts
Plan-first reasoningA pattern where the agent emits a numbered plan before any tool call, then executes it step by step.Goal pinningA pattern where the original goal is auto-injected at the top of every turn so the LLM cannot lose the thread.AI agentAn autonomous program that uses an LLM to plan and execute multi-step tasks via tools, in a loop, until a goal is reached.
Newsletter
Get the next post in your inbox.
Engineering notes from the Digitorn team. No marketing, no launch announcements, no "10 prompts that will change your life". Just the things we write that we'd want to read.