evaluation-rubrics_skill

Overview

This skill provides a practical guide to create, test, and apply evaluation rubrics that produce consistent, transparent, and actionable assessments. It helps teams define observable criteria, choose appropriate scoring scales, write clear performance descriptors, and calibrate reviewers to reduce bias. Use it to set acceptance thresholds, compare alternatives objectively, and give targeted feedback.

How this skill works

The skill walks you through a six-step workflow: define purpose and scope, identify criteria, design the scale, write performance descriptors, test and calibrate, then use and iterate. It includes rubric patterns (analytic, holistic, single-point, checklist, standards-based), guardrails for descriptor clarity and observability, and guidance on weighting and calibration to improve inter-rater reliability. Outputs are a ready-to-use rubric with scoring rules and example anchors.

When to use it

• Designing consistent assessments for code reviews, essays, designs, or proposals
• Creating vendor, candidate, or grant evaluation frameworks for objective comparisons
• Setting quality gates or pass/fail thresholds for product launches or compliance checks
• Improving inter-rater reliability when multiple reviewers must score the same work
• Delivering actionable feedback tied to specific, observable criteria

Best practices

• Limit criteria to 4–8 observable dimensions to balance coverage and usability
• Write concrete, behavior-focused descriptors that distinguish adjacent levels
• Choose scale granularity to match discrimination needs (1-4 or 1-5 are common)
• Include anchor examples for each level and run calibration sessions with reviewers
• Share the rubric with evaluatees before assessment and iterate based on real scoring data

Example use cases

• A code-review rubric scoring Correctness, Readability, Efficiency, and Security on a 1–5 scale
• A vendor-selection checklist with must-have binary gates plus weighted criteria for scoring
• A training certification rubric mapping competencies to Novice/Competent/Expert levels
• A design critique using a single-point rubric to encourage growth-focused written feedback
• Sprint review acceptance criteria with thresholds (e.g., must score ≥4 on Stability to pass)