Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Knowledge library across 2 sources: Influence, How to Win Friends. Use to apply or cross-reference their frameworks on compliance, reciprocity, and framing. Each book has its own references/reference-<slug>.md, loaded on demand.
.claude/skills/ariel-lee-1023-test-over-budget-library/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -26% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 6% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -50% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -52% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -65% | 0% |
<!-- argument-hint: topic, framework name, or book] -->
Books: 2 | Generated: 2026-08-04 | Depth: study
| Book (→ file) | Reach for it when you need… | Its one big idea | |---|---|---| | Influence → references/reference-cialdini-influence.md | to name the lever someone is pulling on you | Compliance runs on six automatic triggers | | How to Win Friends → references/reference-carnegie-persuasion.md | to change someone's mind without resentment | Criticism never survives contact with ego |
Covers these sources only. For a topic no book here addresses, say so rather than inventing it.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 19,741 | 41,986 | +113% | 1 | 1 | 0% | 3,222 | 2,371 | -26% | 0 | 0 | — |
case-02 | fail→pass | 10,428 | 8,081 | -23% | 1 | 1 | 0% | 1,514 | 1,603 | +6% | 0 | 0 | — |
case-03 | fail→pass | 26,523 | 26,516 | -0% | 1 | 1 | 0% | 4,317 | 2,143 | -50% | 0 | 0 | — |
case-04 | pass→pass | 11,972 | 4,389 | -63% | 1 | 1 | 0% | 1,898 | 1,072 | -44% | 0 | 0 | — |
case-05 | fail→pass | 12,470 | 3,914 | -69% | 1 | 1 | 0% | 2,118 | 1,009 | -52% | 0 | 0 | — |
case-06 | fail→fail | 20,788 | 11,979 | -42% | 1 | 1 | 0% | 3,082 | 2,161 | -30% | 0 | 0 | — |
case-07 | fail→pass | 14,616 | 2,696 | -82% | 1 | 1 | 0% | 2,094 | 725 | -65% | 0 | 0 | — |
case-08 | fail→pass | 13,292 | 13,327 | +0% | 1 | 1 | 0% | 1,438 | 2,454 | +71% | 0 | 0 | — |
case-09 | fail→pass | 19,847 | 11,464 | -42% | 1 | 1 | 0% | 3,590 | 2,134 | -41% | 0 | 0 | — |
case-10 | fail→fail | 14,337 | 10,115 | -29% | 1 | 1 | 0% | 2,265 | 1,971 | -13% | 0 | 0 | — |
case-11 | fail→pass | 12,778 | 10,912 | -15% | 1 | 1 | 0% | 2,126 | 1,883 | -11% | 0 | 0 | — |
case-12 | fail→fail | 14,353 | 10,599 | -26% | 1 | 1 | 0% | 1,877 | 1,893 | +1% | 0 | 0 | — |
case-13 | pass→pass | 12,074 | 9,004 | -25% | 1 | 1 | 0% | 1,751 | 885 | -49% | 0 | 0 | — |
case-14 | pass→pass | 11,306 | 5,459 | -52% | 1 | 1 | 0% | 1,630 | 968 | -41% | 0 | 0 | — |
case-15 | fail→fail | 11,826 | 10,362 | -12% | 1 | 1 | 0% | 1,867 | 1,719 | -8% | 0 | 0 | — |
case-16 | fail→pass | 15,606 | 5,175 | -67% | 1 | 1 | 0% | 2,175 | 1,096 | -50% | 0 | 0 | — |
case-17 | fail→pass | 8,131 | 4,257 | -48% | 1 | 1 | 0% | 1,391 | 961 | -31% | 0 | 0 | — |
case-18 | fail→pass | 3,028 | 3,348 | +11% | 1 | 1 | 0% | 398 | 694 | +74% | 0 | 0 | — |
case-19 | fail→pass | 14,739 | 5,226 | -65% | 1 | 1 | 0% | 1,850 | 1,162 | -37% | 0 | 0 | — |
case-20 | fail→fail | 9,788 | 4,552 | -53% | 1 | 1 | 0% | 1,539 | 1,028 | -33% | 0 | 0 | — |
case-21 | fail→pass | 11,875 | 3,235 | -73% | 1 | 1 | 0% | 1,944 | 868 | -55% | 0 | 0 | — |
case-22 | fail→pass | 20,313 | 3,650 | -82% | 1 | 1 | 0% | 2,991 | 798 | -73% | 0 | 0 | — |
case-23 | fail→pass | 12,475 | 3,170 | -75% | 1 | 1 | 0% | 1,790 | 868 | -52% | 0 | 0 | — |
case-24 | pass→pass | 9,699 | 5,591 | -42% | 1 | 1 | 0% | 1,554 | 1,044 | -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. 24 cases were attempted, and 23 counted toward the lift figure. The other 1 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 +63 percentage points is the difference between those two pass rates over the 23 comparable cases.
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.