Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →8 peer review skills. Trigger: reviewing manuscripts, comparing papers, quality assessment. Design: systematic review criteria, evaluation rubrics, and automated review tools.
.claude/skills/brycewang-stanford-paper-review-skills/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-12 | ✗→✓ | ▲ Improved | -53% | 0% |
| case-18 | ✗→✓ | ▲ Improved | -64% | 0% |
| case-21 | ✓→✗ | ▼ Worse | -72% | 0% |
| case-23 | ✓→✗ | ▼ Worse | -74% | 0% |
| case-20 | ✓→✓ | ＝ Same ✓ | 21% | 0% |
Select the skill matching the user's need, then read its SKILL.md.
| Skill | Description | |-------|-------------| | automated-review-guide | AI-assisted peer review tools, workflows, and quality standards | | latte-review-guide | Automate systematic literature reviews with LatteReview AI agents | | paper-critique-framework | Structured framework for writing peer review reports and paper critiques | | paper-reading-assistant | AI-assisted paper reading, PDF Q&A, and summarization workflows | | peer-review-guide | Conduct thorough, constructive peer reviews and evaluate research papers | | rebuttal-writing-guide | Write effective rebuttals to reviewer comments for journal submissions | | review-response-guide | Craft effective point-by-point reviewer response letters | | scientify-write-review-paper | Write literature reviews and survey papers from collected papers |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 9,881 | 4,647 | -53% | 1 | 1 | 0% | 1,529 | 539 | -65% | 0 | 0 | — |
case-02 | fail→fail | 14,657 | 4,635 | -68% | 1 | 1 | 0% | 2,489 | 677 | -73% | 0 | 0 | — |
case-03 | fail→fail | 17,092 | 5,174 | -70% | 1 | 1 | 0% | 2,664 | 628 | -76% | 0 | 0 | — |
case-04 | fail→fail | 17,335 | 4,478 | -74% | 1 | 1 | 0% | 2,947 | 735 | -75% | 0 | 0 | — |
case-05 | fail→fail | 19,272 | 4,336 | -78% | 1 | 1 | 0% | 3,321 | 729 | -78% | 0 | 0 | — |
case-06 | fail→fail | 16,131 | 3,788 | -77% | 1 | 1 | 0% | 2,469 | 705 | -71% | 0 | 0 | — |
case-07 | fail→fail | 24,207 | 5,391 | -78% | 1 | 1 | 0% | 3,601 | 778 | -78% | 0 | 0 | — |
case-08 | fail→fail | 20,506 | 18,699 | -9% | 1 | 1 | 0% | 3,134 | 653 | -79% | 0 | 0 | — |
case-09 | fail→fail | 20,009 | 4,290 | -79% | 1 | 1 | 0% | 3,114 | 696 | -78% | 0 | 0 | — |
case-10 | fail→fail | 16,012 | 3,862 | -76% | 1 | 1 | 0% | 2,470 | 682 | -72% | 0 | 0 | — |
case-11 | fail→fail | 12,103 | 4,019 | -67% | 1 | 1 | 0% | 1,868 | 568 | -70% | 0 | 0 | — |
case-12 | fail→pass | 15,477 | 6,546 | -58% | 1 | 1 | 0% | 2,385 | 1,116 | -53% | 0 | 0 | — |
case-13 | fail→fail | 20,306 | 4,689 | -77% | 1 | 1 | 0% | 3,388 | 765 | -77% | 0 | 0 | — |
case-14 | fail→fail | 20,173 | 4,013 | -80% | 1 | 1 | 0% | 2,807 | 628 | -78% | 0 | 0 | — |
case-15 | fail→fail | 18,447 | 5,323 | -71% | 1 | 1 | 0% | 2,913 | 748 | -74% | 0 | 0 | — |
case-16 | fail→fail | 16,246 | 5,254 | -68% | 1 | 1 | 0% | 2,574 | 678 | -74% | 0 | 0 | — |
case-17 | fail→fail | 15,833 | 3,509 | -78% | 1 | 1 | 0% | 2,573 | 655 | -75% | 0 | 0 | — |
case-18 | fail→pass | 9,893 | 1,835 | -81% | 1 | 1 | 0% | 1,681 | 611 | -64% | 0 | 0 | — |
case-19 | fail→fail | 16,003 | 4,243 | -73% | 1 | 1 | 0% | 2,307 | 647 | -72% | 0 | 0 | — |
case-20 | pass→pass | 14,158 | 14,526 | +3% | 1 | 1 | 0% | 2,325 | 2,814 | +21% | 0 | 0 | — |
case-21 | pass→fail | 18,171 | 5,883 | -68% | 1 | 1 | 0% | 2,861 | 801 | -72% | 0 | 0 | — |
case-22 | pass→pass | 11,158 | 8,947 | -20% | 1 | 1 | 0% | 2,180 | 2,055 | -6% | 0 | 0 | — |
case-23 | pass→fail | 17,474 | 24,880 | +42% | 1 | 1 | 0% | 2,795 | 713 | -74% | 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. 23 cases were attempted, and 4 counted toward the lift figure. The other 19 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of 0 percentage points is the difference between those two pass rates over the 4 comparable cases. 2 cases got worse with the skill loaded, and they are 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.