Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Tactic: Fill a six-slot QALMRI worksheet for one paper: Question, Alternatives, Logic, Method, Results, and Inference. Use for a structured reading worksheet rather than a graded evaluation.
.claude/skills/yogsoth-ai-qalmri-worksheet/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | -72% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -37% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -51% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -81% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -78% | 0% |
paper-fetch; stop on not_found.context/papers/<dir>/qalmri-worksheet/.qalmri with source_path and meta_path.01-qalmri.md with all six sections.Only Inference is judgment. Question, Alternatives, Logic, Method, and Results record what the paper says. Leave unsupported slots empty with a reason rather than inventing content or broadening the method.
Frontmatter records sop: qalmri, tactic: qalmri-worksheet, written_at, and slots_empty. Report the one-sentence question, empty slots, whether the paper states alternatives, and the output path.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-02 | fail→fail | 13,357 | 14,967 | +12% | 1 | 1 | 0% | 1,535 | 497 | -68% | 0 | 0 | — |
case-01 | fail→fail | 14,758 | 14,843 | +1% | 1 | 1 | 0% | 246 | 403 | +64% | 0 | 0 | — |
case-03 | fail→fail | 14,908 | 14,310 | -4% | 1 | 1 | 0% | 232 | 555 | +139% | 0 | 0 | — |
case-04 | pass→fail | 11,242 | 11,103 | -1% | 1 | 1 | 0% | 1,008 | 796 | -21% | 0 | 0 | — |
case-05 | fail→pass | 15,985 | 7,331 | -54% | 1 | 1 | 0% | 1,961 | 556 | -72% | 0 | 0 | — |
case-06 | fail→pass | 13,083 | 9,390 | -28% | 1 | 1 | 0% | 1,369 | 860 | -37% | 0 | 0 | — |
case-07 | fail→pass | 19,058 | 10,375 | -46% | 1 | 1 | 0% | 2,217 | 1,097 | -51% | 0 | 0 | — |
case-08 | fail→pass | 16,605 | 6,419 | -61% | 1 | 1 | 0% | 2,021 | 388 | -81% | 0 | 0 | — |
case-09 | fail→pass | 40,457 | 6,678 | -83% | 1 | 1 | 0% | 2,084 | 458 | -78% | 0 | 0 | — |
case-10 | fail→fail | 14,574 | 6,903 | -53% | 1 | 1 | 0% | 1,514 | 455 | -70% | 0 | 0 | — |
case-11 | pass→fail | 13,566 | 7,585 | -44% | 1 | 1 | 0% | 1,347 | 604 | -55% | 0 | 0 | — |
case-12 | fail→fail | 10,509 | 8,273 | -21% | 1 | 1 | 0% | 924 | 782 | -15% | 0 | 0 | — |
case-13 | fail→pass | 15,123 | 8,047 | -47% | 1 | 1 | 0% | 1,605 | 787 | -51% | 0 | 0 | — |
case-14 | fail→pass | 16,204 | 7,241 | -55% | 1 | 1 | 0% | 1,963 | 571 | -71% | 0 | 0 | — |
case-15 | fail→pass | 14,804 | 7,125 | -52% | 1 | 1 | 0% | 1,928 | 427 | -78% | 0 | 0 | — |
case-16 | pass→pass | 14,592 | 9,028 | -38% | 1 | 1 | 0% | 1,537 | 769 | -50% | 0 | 0 | — |
case-17 | fail→pass | 15,697 | 7,852 | -50% | 1 | 1 | 0% | 1,577 | 614 | -61% | 0 | 0 | — |
case-18 | pass→pass | 13,757 | 7,915 | -42% | 1 | 1 | 0% | 1,750 | 661 | -62% | 0 | 0 | — |
case-19 | fail→pass | 14,544 | 6,668 | -54% | 1 | 1 | 0% | 1,586 | 393 | -75% | 0 | 0 | — |
case-20 | pass→pass | 26,290 | 32,115 | +22% | 1 | 1 | 0% | 3,119 | 4,363 | +40% | 0 | 0 | — |
case-21 | pass→pass | 18,896 | 28,850 | +53% | 1 | 1 | 0% | 2,951 | 4,960 | +68% | 0 | 0 | — |
case-22 | pass→pass | 10,418 | 11,367 | +9% | 1 | 1 | 0% | 1,095 | 1,479 | +35% | 0 | 0 | — |
case-23 | pass→pass | 17,943 | 6,805 | -62% | 1 | 1 | 0% | 2,316 | 479 | -79% | 0 | 0 | — |
case-24 | pass→fail | 10,094 | 7,103 | -30% | 1 | 1 | 0% | 780 | 505 | -35% | 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. 24 cases were attempted, and 19 counted toward the lift figure. The other 5 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 +29 percentage points is the difference between those two pass rates over the 19 comparable cases. 3 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.