Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Run tests and systematically fix all failing tests using smart error grouping. Use when user asks to fix failing tests, mentions test failures, runs test suite and failures occur, or requests to make tests pass.
.claude/skills/davila7-test-fixing/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✗→✓ | ▲ Improved | -21% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 24% | 0% |
| case-19 | ✗→✓ | ▲ Improved | -9% | 0% |
| case-21 | ✓→✗ | ▼ Worse | -24% | 0% |
| case-22 | ✓→✗ | ▼ Worse | -74% | 0% |
Systematically identify and fix all failing tests using smart grouping strategies.
Run make test to identify all failing tests.
Analyze output for:
Group similar failures by:
Prioritize groups by:
For each group (starting with highest impact):
git diffbash uv run pytest tests/path/to/test_file.py -v uv run pytest -k "pattern" -v
Infrastructure first:
Then API changes:
Finally, logic issues:
After all groups fixed:
make testgit diff to understand recent changesUser: "The tests are failing after my refactor"
make test → 15 failures identified| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 3,025 | 4,235 | +40% | 1 | 1 | 0% | 464 | 842 | +81% | 0 | 0 | — |
case-02 | fail→fail | 9,596 | 2,732 | -72% | 1 | 1 | 0% | 1,271 | 885 | -30% | 0 | 0 | — |
case-03 | fail→fail | 2,400 | 3,487 | +45% | 1 | 1 | 0% | 329 | 802 | +144% | 0 | 0 | — |
case-04 | pass→pass | 5,533 | 4,288 | -23% | 1 | 1 | 0% | 1,023 | 1,467 | +43% | 0 | 0 | — |
case-05 | pass→pass | 4,503 | 2,005 | -55% | 1 | 1 | 0% | 955 | 1,084 | +14% | 0 | 0 | — |
case-06 | fail→pass | 7,801 | 2,706 | -65% | 1 | 1 | 0% | 1,463 | 1,155 | -21% | 0 | 0 | — |
case-07 | fail→pass | 3,978 | 1,790 | -55% | 1 | 1 | 0% | 734 | 911 | +24% | 0 | 0 | — |
case-08 | fail→fail | 6,967 | 1,380 | -80% | 1 | 1 | 0% | 1,200 | 879 | -27% | 0 | 0 | — |
case-09 | pass→pass | 11,111 | 3,761 | -66% | 1 | 1 | 0% | 2,013 | 1,299 | -35% | 0 | 0 | — |
case-10 | pass→pass | 6,237 | 3,192 | -49% | 1 | 1 | 0% | 951 | 1,188 | +25% | 0 | 0 | — |
case-11 | pass→pass | 9,719 | 4,245 | -56% | 1 | 1 | 0% | 1,622 | 1,314 | -19% | 0 | 0 | — |
case-12 | pass→pass | 4,099 | 1,245 | -70% | 1 | 1 | 0% | 769 | 824 | +7% | 0 | 0 | — |
case-13 | pass→pass | 7,715 | 3,143 | -59% | 1 | 1 | 0% | 1,304 | 1,211 | -7% | 0 | 0 | — |
case-14 | pass→pass | 7,039 | 2,934 | -58% | 1 | 1 | 0% | 1,239 | 1,180 | -5% | 0 | 0 | — |
case-15 | pass→pass | 8,585 | 5,211 | -39% | 1 | 1 | 0% | 1,506 | 1,467 | -3% | 0 | 0 | — |
case-16 | fail→fail | 5,099 | 1,998 | -61% | 1 | 1 | 0% | 947 | 960 | +1% | 0 | 0 | — |
case-17 | pass→pass | 2,620 | 1,931 | -26% | 1 | 1 | 0% | 417 | 927 | +122% | 0 | 0 | — |
case-18 | pass→pass | 5,225 | 2,268 | -57% | 1 | 1 | 0% | 797 | 1,045 | +31% | 0 | 0 | — |
case-19 | fail→pass | 5,988 | 2,281 | -62% | 1 | 1 | 0% | 1,081 | 989 | -9% | 0 | 0 | — |
case-20 | fail→fail | 3,656 | 3,281 | -10% | 1 | 1 | 0% | 648 | 823 | +27% | 0 | 0 | — |
case-21 | pass→fail | 4,835 | 2,709 | -44% | 1 | 1 | 0% | 1,138 | 866 | -24% | 0 | 0 | — |
case-22 | pass→fail | 14,102 | 2,233 | -84% | 1 | 1 | 0% | 2,931 | 768 | -74% | 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 17 counted toward the lift figure. The other 5 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 +5 percentage points is the difference between those two pass rates over the 17 comparable cases. 2 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.