Prompting Agents

Expert guidance for writing prompts that steer AI agents effectively. Model-agnostic principles with model-specific tips where behavior differs.

Before Writing

Gather this context (ask if not provided):

1. Agent type — Single agent or multi-agent? API, CLI-based, or desktop?
2. Task shape — What does the agent do? What tools does it have?
3. Current problems — What's going wrong? (too verbose, wrong tools, hallucinations, overthinking)
4. Model — Which model? Behavior varies significantly across providers.

Core Principles (Model-Agnostic)

Clarity Over Cleverness