Prompt Engineering / Role Prompting
Match the right persona to a specific task — when to use a teacher vs. coach vs. consultant vs. peer.
Why This Prompt Exists
Most people pick a role arbitrarily — “act as a teacher” — without considering whether that’s the right role for the task.
You get:
- using a teacher when you need a coach (too prescriptive)
- using a consultant when you need a peer (too distant)
- using a critic when you need a collaborator (too negative)
- inconsistent role selection across similar tasks
- suboptimal outcomes because the role doesn’t match the task
But roles have distinct purposes:
- teacher: explains concepts, gives instruction, assumes knowledge gap
- coach: guides, asks questions, builds independence
- consultant: advises, delivers recommendations, expert but external
- peer: collaborates, shares experience, equal footing
- critic: evaluates, identifies flaws, assumes existing work
- facilitator: structures process, enables others
Without mapping, you use the wrong role for the task.
This prompt recommends the optimal role for any task.
The Prompt
Assume the role of a prompt strategist who matches roles to tasks. Your task is to recommend the optimal persona for a given task. Generate: 1. TASK ANALYSIS - Task description - User's current skill level (Novice / Intermediate / Expert) - Desired outcome (Learn / Do / Decide / Improve / Evaluate) 2. ROLE CANDIDATES (ranked by fit) | Role | Fit Score (1-10) | Rationale | |------|------------------|-----------| | Teacher | X/10 | [Why this fits or doesn't] | | Coach | X/10 | [Why this fits or doesn't] | | Consultant | X/10 | [Why this fits or doesn't] | | Peer | X/10 | [Why this fits or doesn't] | | Critic | X/10 | [Why this fits or doesn't] | | Facilitator | X/10 | [Why this fits or doesn't] | 3. RECOMMENDED ROLE - Best role for this task - Runner-up (if user prefers alternative style) 4. WHY THIS ROLE WORKS - How this role's approach matches the task 5. CUSTOMIZED ROLE PROMPT - A ready-to-use prompt with the recommended role 6. WHAT TO AVOID - Roles that would be counterproductive - Why they would fail INPUTS: Task description: [E.G., "Help me debug my Python code"] User's current skill level: [NOVICE / INTERMEDIATE / EXPERT] Desired outcome: [LEARN / DO / DECIDE / IMPROVE / EVALUATE] Previous role used (if any): [E.G., "I tried 'act as a senior developer' but got answers I couldn't understand"] Context: [E.G., "I'm a junior developer, first job"] RULES: - Match role to user skill level (novices need teachers, experts need peers) - Match role to desired outcome (learning → teacher, doing → peer, evaluating → critic) - Avoid over-assigning "consultant" (often too formal and distant) - Consider user preference (some people want direct answers, others want guidance) - Flag if the task doesn't require a role at all (sometimes no role is best)
How To Use It
- Run this before designing any role-based prompt — start with the right role.
- Consider user skill level carefully — a teacher for a novice, a peer for an expert.
- Test multiple roles on the same task — see which produces better outcomes.
- Build a role library for common task types (debugging, brainstorming, learning, deciding).
- Re-map roles as user skill levels change — a learner becomes a peer over time.
Example Input
Task description:
“Help me understand why my machine learning model is overfitting”
User’s current skill level:
“Intermediate — I know the basics but struggle with diagnosis”
Desired outcome:
“LEARN — I want to understand how to diagnose overfitting myself next time”
Previous role used (if any):
“I tried ‘act as a senior ML engineer’ but got too much jargon”
Context:
“Building my first production model”
Why It Works
Most role selection is intuition — “this feels right” — which often picks the wrong role for the task.
This framework improves outcomes by forcing:
- task analysis (what are we actually trying to do?)
- skill level assessment (who is the user?)
- outcome specification (learn, do, decide, improve, evaluate)
- role ranking (which fits best and why)
- avoidance guidance (what roles would hurt)
Great role-to-task mapping doesn’t guess — it matches the role to what the user actually needs.
Build Better AI Systems
Subscribe for advanced prompt engineering, AI coding tools, debugging frameworks, and practical strategies for developers and engineers.