Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Create a new browser QA flow file from the template
.claude/skills/elie222-qa-new-flow/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | -36% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -46% | 0% |
| case-20 | ✗→✓ | ▲ Improved | 25% | 0% |
| case-05 | ✓→✗ | ▼ Worse | -48% | 0% |
| case-16 | ✓→✗ | ▼ Worse | -42% | 0% |
You are creating a new browser QA flow spec in qa/browser-flows.
Args: $ARGUMENTS
If no args or --help is present, print usage and stop.
Usage:
/qa-new-flow --id=flow-id --title="Short title" --resources=assistant-settings,conversation-rules --goal="What it verifies"--parallel-safe=true --conflicts-with=other-flow-id,another-flow-id --preconditions="Signed in" --cleanup="Remove test rule"Steps:
id, title, resources).id is a URL-safe slug (lowercase, numbers, dashes only) and matches the filename.qa/browser-flows/<id>.md using qa/browser-flows/_template.md as a base.parallel_safe, conflicts_with), include them in the front matter.conflicts_with as a YAML list by splitting the --conflicts-with value on commas (even for a single id).--goal is provided, replace the Goal section placeholder with it.--preconditions is provided, replace the existing Preconditions section placeholder list with those items.--cleanup is provided, replace the Cleanup section placeholder with it.Do not overwrite an existing flow file without explicit confirmation.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 9,324 | 5,806 | -38% | 1 | 1 | 0% | 1,793 | 792 | -56% | 0 | 0 | — |
case-02 | fail→fail | 6,623 | 6,704 | +1% | 1 | 1 | 0% | 1,123 | 683 | -39% | 0 | 0 | — |
case-03 | fail→fail | 11,237 | 5,133 | -54% | 1 | 1 | 0% | 2,186 | 684 | -69% | 0 | 0 | — |
case-04 | fail→fail | 6,168 | 8,370 | +36% | 1 | 1 | 0% | 327 | 1,907 | +483% | 0 | 0 | — |
case-05 | pass→fail | 14,984 | 6,902 | -54% | 1 | 1 | 0% | 3,284 | 1,712 | -48% | 0 | 0 | — |
case-16 | pass→fail | 8,217 | 6,782 | -17% | 1 | 1 | 0% | 1,316 | 766 | -42% | 0 | 0 | — |
case-06 | pass→fail | 14,660 | 7,003 | -52% | 1 | 1 | 0% | 2,453 | 674 | -73% | 0 | 0 | — |
case-07 | fail→pass | 9,435 | 6,990 | -26% | 1 | 1 | 0% | 1,546 | 992 | -36% | 0 | 0 | — |
case-08 | fail→fail | 8,654 | 8,287 | -4% | 1 | 1 | 0% | 1,649 | 782 | -53% | 0 | 0 | — |
case-09 | fail→fail | 7,542 | 5,781 | -23% | 1 | 1 | 0% | 1,397 | 684 | -51% | 0 | 0 | — |
case-10 | fail→fail | 11,934 | 6,169 | -48% | 1 | 1 | 0% | 1,996 | 675 | -66% | 0 | 0 | — |
case-11 | fail→fail | 6,194 | 6,545 | +6% | 1 | 1 | 0% | 1,124 | 817 | -27% | 0 | 0 | — |
case-12 | pass→fail | 10,712 | 7,455 | -30% | 1 | 1 | 0% | 2,049 | 665 | -68% | 0 | 0 | — |
case-13 | pass→fail | 8,976 | 5,714 | -36% | 1 | 1 | 0% | 1,651 | 660 | -60% | 0 | 0 | — |
case-14 | fail→pass | 9,244 | 2,943 | -68% | 1 | 1 | 0% | 1,677 | 910 | -46% | 0 | 0 | — |
case-15 | pass→pass | 7,747 | 3,432 | -56% | 1 | 1 | 0% | 1,172 | 822 | -30% | 0 | 0 | — |
case-17 | pass→fail | 9,751 | 7,869 | -19% | 1 | 1 | 0% | 1,804 | 776 | -57% | 0 | 0 | — |
case-18 | fail→fail | 7,849 | 5,848 | -25% | 1 | 1 | 0% | 1,537 | 653 | -58% | 0 | 0 | — |
case-19 | pass→fail | 5,360 | 6,016 | +12% | 1 | 1 | 0% | 775 | 798 | +3% | 0 | 0 | — |
case-20 | fail→pass | 5,141 | 3,613 | -30% | 1 | 1 | 0% | 781 | 979 | +25% | 0 | 0 | — |
case-21 | fail→fail | 13,284 | 6,544 | -51% | 1 | 1 | 0% | 2,378 | 728 | -69% | 0 | 0 | — |
case-22 | pass→fail | 8,235 | 6,866 | -17% | 1 | 1 | 0% | 1,517 | 714 | -53% | 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 5 counted toward the lift figure. The other 17 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 -23 percentage points is the difference between those two pass rates over the 5 comparable cases. 9 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.