Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Create a clear grading rubric with criteria and performance-level descriptors that make scoring fair, fast, and consistent. Use when asked to build a rubric, create grading criteria, design an assessment scoring guide, or make grading more objective. Produces an analytic rubric table (criteria × performance levels) with concrete, observable descriptors and a points scheme — plus a short version students can self-check against.
.claude/skills/mohitagw15856-rubric-builder/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -22% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 63% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 33% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 88% | 0% |
A good rubric turns "this feels like a B" into a defensible, repeatable judgment — and tells students exactly how to do better. This skill builds one with observable descriptors at each level.
Given the assignment and level, build the full rubric anyway, inferring sensible criteria and weighting. Mark anything assumed. Never leave "describe level]"; write concrete descriptors.
Ask for (if not already provided):
| Criterion (weight) | Exemplary (4) | Proficient (3) | Developing (2) | Beginning (1) | |---|---|---|---|---| | Criterion 1] | observable descriptor | | | | | Criterion 2] | | | | | | Criterion 3] | | | | |
Each cell describes what the work actually looks like at that level — observable evidence, not "excellent/good/poor."
How levels convert to points/grade, including weighting.
A short "before you submit, check you've…" version students can self-assess against.
2–3 sentence starters per criterion to speed up consistent written feedback.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 36,158 | 26,970 | -25% | 1 | 1 | 0% | 5,605 | 4,395 | -22% | 0 | 0 | — |
case-02 | fail→fail | 35,335 | 21,857 | -38% | 1 | 1 | 0% | 4,817 | 4,362 | -9% | 0 | 0 | — |
case-03 | fail→fail | 39,641 | 32,733 | -17% | 1 | 1 | 0% | 4,943 | 4,332 | -12% | 0 | 0 | — |
case-04 | fail→fail | 10,582 | 9,699 | -8% | 1 | 1 | 0% | 792 | 1,237 | +56% | 0 | 0 | — |
case-05 | pass→pass | 23,280 | 27,530 | +18% | 1 | 1 | 0% | 3,383 | 4,643 | +37% | 0 | 0 | — |
case-06 | pass→fail | 24,441 | 39,313 | +61% | 1 | 1 | 0% | 3,233 | 4,644 | +44% | 0 | 0 | — |
case-07 | fail→pass | 25,573 | 29,898 | +17% | 1 | 1 | 0% | 4,005 | 4,719 | +18% | 0 | 0 | — |
case-08 | fail→pass | 23,308 | 29,133 | +25% | 1 | 1 | 0% | 2,727 | 4,452 | +63% | 0 | 0 | — |
case-09 | fail→pass | 18,148 | 14,473 | -20% | 1 | 1 | 0% | 2,363 | 3,149 | +33% | 0 | 0 | — |
case-10 | fail→pass | 20,513 | 24,591 | +20% | 1 | 1 | 0% | 1,994 | 3,745 | +88% | 0 | 0 | — |
case-11 | fail→fail | 15,912 | 24,414 | +53% | 1 | 1 | 0% | 2,658 | 3,819 | +44% | 0 | 0 | — |
case-12 | fail→fail | 24,348 | 30,505 | +25% | 1 | 1 | 0% | 4,419 | 4,283 | -3% | 0 | 0 | — |
case-13 | fail→pass | 14,809 | 21,616 | +46% | 1 | 1 | 0% | 1,615 | 3,399 | +110% | 0 | 0 | — |
case-14 | pass→pass | 28,665 | 29,845 | +4% | 1 | 1 | 0% | 3,680 | 3,976 | +8% | 0 | 0 | — |
case-15 | fail→pass | 27,712 | 34,067 | +23% | 1 | 1 | 0% | 3,317 | 4,674 | +41% | 0 | 0 | — |
case-16 | fail→pass | 23,406 | 25,353 | +8% | 1 | 1 | 0% | 2,799 | 3,927 | +40% | 0 | 0 | — |
case-17 | fail→fail | 24,679 | 27,611 | +12% | 1 | 1 | 0% | 2,722 | 4,388 | +61% | 0 | 0 | — |
case-18 | fail→fail | 27,334 | 24,232 | -11% | 1 | 1 | 0% | 2,771 | 3,730 | +35% | 0 | 0 | — |
case-19 | fail→pass | 56,830 | 42,161 | -26% | 1 | 1 | 0% | 7,332 | 6,407 | -13% | 0 | 0 | — |
case-20 | fail→pass | 16,327 | 24,660 | +51% | 1 | 1 | 0% | 1,934 | 3,192 | +65% | 0 | 0 | — |
case-21 | fail→pass | 17,627 | 29,863 | +69% | 1 | 1 | 0% | 2,048 | 4,053 | +98% | 0 | 0 | — |
case-22 | fail→pass | 39,936 | 35,054 | -12% | 1 | 1 | 0% | 5,926 | 4,802 | -19% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted. The headline lift of +50 percentage points is the difference between those two pass rates over the 22 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.