Rubric Prompting: Score Outputs and Improve Systematically

Rubric Prompting: Score Outputs and Improve Systematically

This lesson is part of Prompt Patterns & Frameworks. You’ll learn this like you would learn a real skill at work: understand the idea, apply a repeatable pattern, then test it on messy inputs.

What you’ll be able to do after this

• Explain the concept in simple language (no unnecessary theory).
• Use a repeatable prompt/workflow pattern you can reuse later.
• Spot common failure modes and fix them with a small, targeted prompt change.

The core idea (explained like a teammate)

Prompt engineering is basically context + constraints + structure. When outputs look “random”, it’s usually because one of those three is weak. Your job is to make the task unambiguous and the output shape predictable.

A reusable pattern

1. Goal: one line — what success looks like.
2. Context: minimum facts needed (not the whole story).
3. Constraints: do/don’t rules, tone, length, format.
4. Output template: fixed sections so the model can’t wander.
5. Verification: a small checklist before final output.

Copy‑paste prompt template

Role: You are a practical expert in Prompt Patterns & Frameworks.
Task: Rubric Prompting: Score Outputs and Improve Systematically.

Rules:
- Ask up to 2 questions if key details are missing.
- Follow the output template exactly.
- If unsure, say you’re unsure (do not invent facts).

Output template:
1) Summary (3 lines)
2) Step-by-step approach
3) Common mistakes (bullets)
4) Mini exercise

Common mistakes (and how to fix them)

• Vague instructions: replace “good/best” with measurable constraints.
• No structure: add an output template with fixed headings.
• Too many goals at once: split into steps (outline → draft → review).
• Blind trust: add a verification checklist for correctness and format.

Mini exercise

Pick a real task you do weekly (SEO blog, support reply, SQL insert, debugging). Use the template above and test it on 3 inputs: one normal, one messy, and one edge case. Note what failed — that note becomes your next prompt improvement.

Difficulty: Intermediate. If it feels heavy, repeat the mini exercise twice — that’s where the skill actually sticks.