Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →E2E test fixture — calls a tool that always fails, used to verify tool_call_error events.
.claude/skills/skrun-dev-audit-fixture-tool-error/SKILL.md| Model | Eval pass | Runs |
|---|---|---|
| gemini-3.6-flash | 100% | 15 |
| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -75% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -10% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -74% | 0% |
| case-09 | ✓→✗ | ▼ Worse | -80% | 0% |
| case-11 | ✓→✗ | ▼ Worse | -72% | 0% |
You are an E2E test fixture. Your only job is to invoke the always_fails tool exactly once with {} as input. The tool will fail with a non-zero exit. After the failure, return a single JSON object: {"result": "tool failed as expected"}.
Do not invoke always_fails more than once. Do not invoke any other tool. Do not retry. Do not attempt recovery.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 6,115 | 3,944 | -36% | 1 | 1 | 0% | 914 | 301 | -67% | 0 | 0 | — |
case-02 | fail→fail | 6,552 | 2,857 | -56% | 1 | 1 | 0% | 982 | 292 | -70% | 0 | 0 | — |
case-03 | fail→fail | 12,274 | 2,213 | -82% | 1 | 1 | 0% | 1,399 | 250 | -82% | 0 | 0 | — |
case-04 | fail→pass | 11,072 | 5,608 | -49% | 1 | 1 | 0% | 1,629 | 401 | -75% | 0 | 0 | — |
case-05 | fail→fail | 14,060 | 3,888 | -72% | 1 | 1 | 0% | 1,947 | 337 | -83% | 0 | 0 | — |
case-06 | fail→fail | 11,523 | 4,988 | -57% | 1 | 1 | 0% | 1,612 | 783 | -51% | 0 | 0 | — |
case-12 | fail→fail | 14,778 | 5,396 | -63% | 1 | 1 | 0% | 2,487 | 412 | -83% | 0 | 0 | — |
case-07 | fail→pass | 9,876 | 8,299 | -16% | 1 | 1 | 0% | 1,531 | 1,376 | -10% | 0 | 0 | — |
case-08 | fail→fail | 10,085 | 4,780 | -53% | 1 | 1 | 0% | 1,401 | 496 | -65% | 0 | 0 | — |
case-09 | pass→fail | 12,255 | 3,053 | -75% | 1 | 1 | 0% | 1,716 | 335 | -80% | 0 | 0 | — |
case-10 | fail→fail | 10,417 | 3,914 | -62% | 1 | 1 | 0% | 1,637 | 381 | -77% | 0 | 0 | — |
case-11 | pass→fail | 13,370 | 4,744 | -65% | 1 | 1 | 0% | 1,833 | 507 | -72% | 0 | 0 | — |
case-13 | fail→fail | 9,914 | 4,726 | -52% | 1 | 1 | 0% | 1,404 | 443 | -68% | 0 | 0 | — |
case-14 | fail→pass | 9,808 | 1,932 | -80% | 1 | 1 | 0% | 1,400 | 366 | -74% | 0 | 0 | — |
case-15 | fail→fail | 16,041 | 3,554 | -78% | 1 | 1 | 0% | 2,284 | 325 | -86% | 0 | 0 | — |
case-16 | fail→fail | 7,468 | 4,566 | -39% | 1 | 1 | 0% | 1,182 | 400 | -66% | 0 | 0 | — |
case-17 | fail→fail | 4,688 | 2,739 | -42% | 1 | 1 | 0% | 592 | 260 | -56% | 0 | 0 | — |
case-18 | fail→fail | 10,385 | 3,545 | -66% | 1 | 1 | 0% | 1,482 | 363 | -76% | 0 | 0 | — |
case-19 | fail→fail | 9,548 | 2,826 | -70% | 1 | 1 | 0% | 1,410 | 284 | -80% | 0 | 0 | — |
case-20 | pass→fail | 13,690 | 4,341 | -68% | 1 | 1 | 0% | 2,003 | 406 | -80% | 0 | 0 | — |
case-21 | pass→fail | 18,124 | 3,426 | -81% | 1 | 1 | 0% | 2,858 | 345 | -88% | 0 | 0 | — |
case-22 | pass→fail | 8,898 | 2,814 | -68% | 1 | 1 | 0% | 1,517 | 292 | -81% | 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, and 4 counted toward the lift figure. The other 18 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 -9 percentage points is the difference between those two pass rates over the 4 comparable cases. 5 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.