Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Greet a caller and echo back a short, structured summary of their message.
.claude/skills/bingzi-233-hello/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -23% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -4% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 14% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -56% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -40% | 0% |
A minimal example skill. It proves the round trip: HTTP request → agent run → schema-validated JSON.
name.message is provided, write a one-line summary of it. Otherwise set summary to an empty string.json{ "greeting": "Hello, <name>!", "summary": "<one-line summary or empty string>" }
This skill needs no tools beyond the read-only sandbox. To let a skill write files or run commands, drop a .mcp.json sidecar next to this file (see the README, "MCP & tools").
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 2,493 | 1,253 | -50% | 1 | 1 | 0% | 528 | 408 | -23% | 0 | 0 | — |
case-02 | fail→pass | 2,138 | 1,367 | -36% | 1 | 1 | 0% | 376 | 362 | -4% | 0 | 0 | — |
case-03 | fail→pass | 2,494 | 1,974 | -21% | 1 | 1 | 0% | 493 | 563 | +14% | 0 | 0 | — |
case-04 | pass→fail | 11,067 | 4,188 | -62% | 1 | 1 | 0% | 2,611 | 1,016 | -61% | 0 | 0 | — |
case-05 | pass→pass | 2,892 | 12,202 | +322% | 1 | 1 | 0% | 717 | 2,678 | +274% | 0 | 0 | — |
case-06 | pass→pass | 4,565 | 4,565 | 0% | 1 | 1 | 0% | 1,078 | 1,176 | +9% | 0 | 0 | — |
case-07 | pass→pass | 1,974 | 1,908 | -3% | 1 | 1 | 0% | 402 | 528 | +31% | 0 | 0 | — |
case-08 | fail→pass | 4,510 | 1,322 | -71% | 1 | 1 | 0% | 887 | 388 | -56% | 0 | 0 | — |
case-09 | fail→pass | 5,160 | 2,345 | -55% | 1 | 1 | 0% | 1,015 | 613 | -40% | 0 | 0 | — |
case-10 | pass→pass | 3,210 | 2,036 | -37% | 1 | 1 | 0% | 599 | 564 | -6% | 0 | 0 | — |
case-11 | fail→pass | 5,125 | 1,763 | -66% | 1 | 1 | 0% | 942 | 482 | -49% | 0 | 0 | — |
case-12 | pass→pass | 2,320 | 1,838 | -21% | 1 | 1 | 0% | 431 | 482 | +12% | 0 | 0 | — |
case-13 | fail→pass | 4,205 | 3,229 | -23% | 1 | 1 | 0% | 966 | 663 | -31% | 0 | 0 | — |
case-14 | fail→pass | 3,510 | 1,612 | -54% | 1 | 1 | 0% | 670 | 436 | -35% | 0 | 0 | — |
case-15 | fail→fail | 5,029 | 1,924 | -62% | 1 | 1 | 0% | 912 | 506 | -45% | 0 | 0 | — |
case-16 | fail→pass | 3,620 | 1,816 | -50% | 1 | 1 | 0% | 802 | 489 | -39% | 0 | 0 | — |
case-17 | fail→pass | 2,630 | 1,353 | -49% | 1 | 1 | 0% | 537 | 353 | -34% | 0 | 0 | — |
case-18 | fail→pass | 3,310 | 1,700 | -49% | 1 | 1 | 0% | 707 | 486 | -31% | 0 | 0 | — |
case-19 | fail→pass | 3,535 | 2,069 | -41% | 1 | 1 | 0% | 765 | 566 | -26% | 0 | 0 | — |
case-20 | fail→pass | 4,801 | 1,263 | -74% | 1 | 1 | 0% | 826 | 383 | -54% | 0 | 0 | — |
case-21 | fail→pass | 3,435 | 1,913 | -44% | 1 | 1 | 0% | 707 | 530 | -25% | 0 | 0 | — |
case-22 | fail→pass | 4,315 | 1,852 | -57% | 1 | 1 | 0% | 930 | 499 | -46% | 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 +64 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.