F68 AI

Library · System Prompt · 5.17

SYSTEM PROMPT CHECKLIST

1 min read

Before putting the Agent into use:

Objective

  • What does the Agent exist to do?

Scope

  • What is it allowed to do?
  • What is it not allowed to do?

Workflow

  • Is the sequence clear?
  • Which steps are mandatory?

Rules

  • Are they short and checkable?

Knowledge

  • Does the Agent know when to use the KB?

Tools

  • Is there a rule for when to use a Tool?

Sources

  • Is there a Source Policy if the job needs research?

Output

  • Does the final result have a clear structure?

Failure

  • What to do when data is missing?
  • What to do when a Tool fails?

Conflict

  • Which Rule takes priority?

If these parts are clear, the Agent has an Instruction foundation good enough to start testing.

SUMMARY

A good System Prompt doesn’t need to be long.

It needs to make clear:

What the Agent does. What it does not do. In what sequence it works. How it uses Knowledge and Tools. What the result must contain. And when it must stop.