AI Agent Prompting Framework

1. Start by giving your AI agent a clear and specific role. 2. Tell the agent exactly what expertise and responsibilities it should have. 3. Define one clear goal instead of giving it a vague objective. 4. Make the goal specific, measurable, actionable, and outcome-focused. 5. Specify exactly what inputs the agent will receive and work with. 6. Inputs can include documents, URLs, databases, emails, PDFs, spreadsheets, and messages. 7. Define which tools the agent is allowed to use to complete its work. 8. Give it access to relevant tools like search, email, calendars, APIs, CRMs, and databases. 9. Map out the workflow so the agent knows what steps to follow. 10. A simple workflow can be analyze → research → plan → execute → verify → deliver. 11. Add decision rules that tell the agent what to do when things get unclear. 12. Tell it to ask questions when information is missing instead of making things up. 13. Define what information the agent should remember for future tasks. 14. Memory can include preferences, previous actions, conversations, and historical outputs. 15. Specify the exact output format you want before the agent starts working. 16. Ask for structured outputs like summaries, recommendations, action plans, or next steps. 17. Add a verification step so the agent checks its own work before delivering it. 18. Tell it to verify accuracy, completeness, calculations, duplicates, and goal achievement. 19. Avoid vague goals, undefined tools, missing workflows, unlimited autonomy, and missing guardrails. 20. The best formula is Role + Goal + Inputs + Tools + Workflow + Rules + Memory + Output + Verification.