Skip to main content

System Prompts

The system prompt defines an agent's personality, capabilities, knowledge, and constraints. It is prepended to every conversation as the first message.

Vault file injection​

Reference vault files directly in the system prompt using @[path/to/file]:

You are a customer support agent for Acme Corp.

Company knowledge base:
@[knowledge/faq.md]
@[knowledge/pricing.md]
@[knowledge/policies.md]

When the agent starts, Synapse reads the referenced vault files and injects their content into the system prompt. Files are re-read each session so updates take effect immediately.

State references in orchestrations​

Within orchestration prompt_template fields, reference shared state values using {state.key}:

You are a writer. Summarise the following research:

{state.research_results}

Write a 500-word article in markdown format.

Tips for effective system prompts​

Be specific about output format​

Always respond with JSON in this exact format:
{
  "summary": "...",
  "confidence": 0.0-1.0,
  "sources": ["url1", "url2"]
}

Define tool usage policy​

- Use the browser tool for any question about current events or specific URLs
- Use the vault only to save final outputs — never intermediate steps
- Use the SQL tool only for SELECT queries unless the user explicitly asks to modify data

Set persona and tone​

You are a concise technical writer. Avoid filler words.
Prefer bullet points over paragraphs. Use markdown headers.
Always include a "Next steps" section at the end.

Constrain scope​

You only answer questions about our software product.
For anything outside that scope, politely redirect the user.

Role and context​

You are a senior analyst at a hedge fund. You have access to our market data database.
The current date is always available via the time tool.
Always quantify uncertainty — never present uncertain information as fact.

System prompt length​

System prompts can be as long as needed, but very long prompts consume input tokens on every turn. For large knowledge bases, store content in the vault and reference it via @[file] injection instead of pasting it directly.

Testing system prompts​

Use the Generate Prompt button in the agent editor to get an AI-generated starting point, then refine it by chatting with the agent. The UI shows each tool call and reasoning step so you can see how the agent interprets the prompt.

Last updated on