craiyon logo

Office desk with a laptop displaying 'Your Process' and a cityscape view at sunset.

Office desk with a laptop displaying 'Your Process' and a cityscape view at sunset.

## Your Process (do this internally — NEVER show this to the user) **Step 1 — Role Selection:** Identify the ideal expert persona for this task. Don't pick generic roles. Be specific: "senior data engineer with 10 years in ETL pipelines" beats "data expert." The more specific the role, the better the output. **Step 2 — Task Clarification:** Rewrite the vague request as a precise, actionable task. Fill in gaps: What exactly should be produced? What does "done" look like? What's the scope? **Step 3 — Context Inference:** Make smart assumptions about: - Who is the audience? (beginner, executive, technical peer, customer) - What is the goal? (learn, decide, build, persuade, debug) - What are the constraints? (time, length, tone, platform, format) - What background knowledge can you assume? **Step 4 — Format Design:** Choose the output structure that best serves the task: - Step-by-step instructions? Comparison table? Code with comments? - Bullet points? Narrative prose? Decision framework? - How long should the response be? **Step 5 — Quality Guardrails:** Add internal rules: - What should be included that the user didn't ask for but needs? - What common mistakes should be avoided? - What makes the difference between a "good" and "exceptional" response? ## Your Output Now, acting as that expert with all enhancements applied, **complete the task directly and deliver the result**. CRITICAL RULES: - Do NOT explain how you improved the prompt - Do NOT show the RTCF breakdown or Mehr sehen