Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when the target venue's submission requirements matter — checking a draft against NeurIPS, ICML or ICLR expectations, deciding which required sections (checklist, broader impact, reproducibility, LLM disclosure) the paper needs, or running the Stage 08 submission-readiness review.
.claude/skills/tangxiangru-venue-checklist/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-22 | ✗→✓ | ▲ Improved | 33% | 0% |
| case-05 | ✓→✗ | ▼ Worse | -18% | 0% |
| case-08 | ✓→✗ | ▼ Worse | -1% | 0% |
| case-04 | ✓→✓ | = Same ✓ | 1% | 0% |
| case-18 | ✓→✓ | = Same ✓ | -6% | 0% |
Read this when the target venue is set in Stage 03, when locking the Stage 07 outline, and again during the Stage 08 readiness review. The run's venue is in run_config.json and is rendered into every stage prompt under Run Configuration.
Across these venues, the following are usually expected:
Planning implications:
Final-check implications:
Planning implications:
Final-check implications:
Planning implications:
Final-check implications:
Before submission, verify:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 25,426 | 16,217 | -36% | 1 | 1 | 0% | 3,346 | 2,435 | -27% | 0 | 0 | — |
case-02 | fail→fail | 11,694 | 15,865 | +36% | 1 | 1 | 0% | 1,194 | 2,356 | +97% | 0 | 0 | — |
case-03 | fail→fail | 19,520 | 14,021 | -28% | 1 | 1 | 0% | 2,305 | 2,014 | -13% | 0 | 0 | — |
case-04 | pass→pass | 23,449 | 13,890 | -41% | 1 | 1 | 0% | 2,812 | 2,848 | +1% | 0 | 0 | — |
case-05 | pass→fail | 21,529 | 14,888 | -31% | 1 | 1 | 0% | 2,677 | 2,199 | -18% | 0 | 0 | — |
case-06 | fail→fail | 18,924 | 15,079 | -20% | 1 | 1 | 0% | 2,930 | 3,122 | +7% | 0 | 0 | — |
case-07 | fail→fail | 25,827 | 20,901 | -19% | 1 | 1 | 0% | 3,300 | 3,639 | +10% | 0 | 0 | — |
case-08 | pass→fail | 16,272 | 7,895 | -51% | 1 | 1 | 0% | 1,721 | 1,702 | -1% | 0 | 0 | — |
case-09 | fail→fail | 14,501 | 13,689 | -6% | 1 | 1 | 0% | 1,540 | 1,905 | +24% | 0 | 0 | — |
case-10 | fail→fail | 18,079 | 10,129 | -44% | 1 | 1 | 0% | 2,084 | 2,339 | +12% | 0 | 0 | — |
case-11 | fail→fail | 22,870 | 22,202 | -3% | 1 | 1 | 0% | 2,657 | 2,987 | +12% | 0 | 0 | — |
case-12 | fail→fail | 24,861 | 33,648 | +35% | 1 | 1 | 0% | 3,441 | 5,121 | +49% | 0 | 0 | — |
case-13 | fail→fail | 17,206 | 22,826 | +33% | 1 | 1 | 0% | 2,681 | 3,485 | +30% | 0 | 0 | — |
case-14 | fail→fail | 17,395 | 20,826 | +20% | 1 | 1 | 0% | 2,793 | 3,133 | +12% | 0 | 0 | — |
case-15 | fail→fail | 14,010 | 15,347 | +10% | 1 | 1 | 0% | 2,047 | 2,195 | +7% | 0 | 0 | — |
case-16 | fail→fail | 20,610 | 12,158 | -41% | 1 | 1 | 0% | 2,560 | 2,568 | +0% | 0 | 0 | — |
case-17 | fail→fail | 14,521 | 6,765 | -53% | 1 | 1 | 0% | 2,042 | 1,738 | -15% | 0 | 0 | — |
case-18 | pass→pass | 17,858 | 6,559 | -63% | 1 | 1 | 0% | 1,928 | 1,806 | -6% | 0 | 0 | — |
case-19 | pass→pass | 17,386 | 29,164 | +68% | 1 | 1 | 0% | 1,948 | 4,089 | +110% | 0 | 0 | — |
case-20 | pass→pass | 16,334 | 19,552 | +20% | 1 | 1 | 0% | 2,300 | 2,757 | +20% | 0 | 0 | — |
case-21 | fail→fail | 17,450 | 17,122 | -2% | 1 | 1 | 0% | 2,075 | 2,531 | +22% | 0 | 0 | — |
case-22 | fail→pass | 22,631 | 24,361 | +8% | 1 | 1 | 0% | 2,578 | 3,436 | +33% | 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 -33 percentage points is the difference between those two pass rates over the 22 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.