Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Surface changed git diff lines that are not covered by tests
.claude/skills/hashgraph-online-test-gap/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-12 | ✗→✓ | ▲ Improved | 415% | 0% |
| case-16 | ✗→✓ | ▲ Improved | -56% | 0% |
| case-21 | ✗→✓ | ▲ Improved | 2% | 0% |
| case-22 | ✗→✓ | ▲ Improved | -56% | 0% |
| case-04 | ✓→✗ | ▼ Worse | 2% | 0% |
Run python3 scripts/gap.py "$@" from the test-gap plugin root.
Purpose: compare the current git diff against the selected base branch, parse the selected or auto-detected coverage report, and report changed lines that are not covered by tests.
Inputs: optional --base, --report, and --format arguments supplied by the user.
Boundaries: do not edit files, do not install packages, do not use the network, and do not hide failures.
Output contract: return either the helper output or a concise failure summary with command, exit status, and next step.
Verification contract: accept success only when the helper exits 0 and prints JSON or markdown matching the requested format.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-03 | fail→fail | 10,512 | 11,682 | +11% | 1 | 1 | 0% | 1,175 | 539 | -54% | 0 | 0 | — |
case-01 | fail→fail | 16,551 | 11,967 | -28% | 1 | 1 | 0% | 1,886 | 561 | -70% | 0 | 0 | — |
case-02 | fail→fail | 7,895 | 16,263 | +106% | 1 | 1 | 0% | 1,325 | 529 | -60% | 0 | 0 | — |
case-04 | pass→fail | 11,529 | 5,598 | -51% | 1 | 1 | 0% | 1,187 | 1,214 | +2% | 0 | 0 | — |
case-05 | fail→fail | 7,299 | 16,957 | +132% | 1 | 1 | 0% | 421 | 2,598 | +517% | 0 | 0 | — |
case-06 | pass→pass | 11,038 | 8,015 | -27% | 1 | 1 | 0% | 1,130 | 661 | -42% | 0 | 0 | — |
case-07 | fail→fail | 7,919 | 11,463 | +45% | 1 | 1 | 0% | 1,351 | 391 | -71% | 0 | 0 | — |
case-08 | fail→fail | 3,393 | 10,235 | +202% | 1 | 1 | 0% | 367 | 495 | +35% | 0 | 0 | — |
case-09 | fail→fail | 9,355 | 10,119 | +8% | 1 | 1 | 0% | 1,659 | 456 | -73% | 0 | 0 | — |
case-10 | pass→fail | 16,458 | 10,715 | -35% | 1 | 1 | 0% | 2,439 | 552 | -77% | 0 | 0 | — |
case-11 | pass→pass | 12,736 | 13,529 | +6% | 1 | 1 | 0% | 2,212 | 1,873 | -15% | 0 | 0 | — |
case-12 | fail→pass | 12,490 | 38,562 | +209% | 1 | 1 | 0% | 1,234 | 6,357 | +415% | 0 | 0 | — |
case-13 | fail→fail | 10,043 | 17,741 | +77% | 1 | 1 | 0% | 759 | 706 | -7% | 0 | 0 | — |
case-14 | fail→fail | 4,007 | 2,538 | -37% | 1 | 1 | 0% | 647 | 504 | -22% | 0 | 0 | — |
case-15 | fail→fail | 9,750 | 15,377 | +58% | 1 | 1 | 0% | 1,771 | 449 | -75% | 0 | 0 | — |
case-16 | fail→pass | 13,947 | 2,836 | -80% | 1 | 1 | 0% | 1,515 | 663 | -56% | 0 | 0 | — |
case-17 | pass→pass | 4,298 | 9,386 | +118% | 1 | 1 | 0% | 763 | 972 | +27% | 0 | 0 | — |
case-18 | fail→fail | 17,544 | 8,047 | -54% | 1 | 1 | 0% | 2,206 | 530 | -76% | 0 | 0 | — |
case-19 | pass→pass | 2,538 | 2,149 | -15% | 1 | 1 | 0% | 440 | 391 | -11% | 0 | 0 | — |
case-20 | fail→fail | 11,854 | 14,682 | +24% | 1 | 1 | 0% | 1,062 | 405 | -62% | 0 | 0 | — |
case-21 | fail→pass | 15,928 | 9,907 | -38% | 1 | 1 | 0% | 1,886 | 1,920 | +2% | 0 | 0 | — |
case-22 | fail→pass | 16,244 | 8,524 | -48% | 1 | 1 | 0% | 1,869 | 818 | -56% | 0 | 0 | — |
case-23 | fail→fail | 31,880 | 10,391 | -67% | 1 | 1 | 0% | 6,908 | 496 | -93% | 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. 23 cases were attempted, and 10 counted toward the lift figure. The other 13 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 10 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.