Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when encountering any bug, test failure, or unexpected behavior, before proposing fixes
.claude/skills/bilal140202-systematic-debugging/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | -46% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -21% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -4% | 0% |
| case-03 | ✓→✓ | ＝ Same ✓ | -17% | 0% |
| case-05 | ✓→✓ | ＝ Same ✓ | -7% | 0% |
The core of systematic debugging is: understand first, then fix. When encountering a bug, clarify symptoms, reproduction conditions, and blast radius. Draw hypotheses, verify them with experiments to isolate the root cause, and only then submit a minimal, verifiable fix.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 15,542 | 14,481 | -7% | 1 | 1 | 0% | 2,637 | 2,775 | +5% | 0 | 0 | — |
case-02 | fail→pass | 12,445 | 5,256 | -58% | 1 | 1 | 0% | 2,334 | 1,259 | -46% | 0 | 0 | — |
case-03 | pass→pass | 10,717 | 6,067 | -43% | 1 | 1 | 0% | 1,784 | 1,474 | -17% | 0 | 0 | — |
case-04 | fail→fail | 14,163 | 12,835 | -9% | 1 | 1 | 0% | 2,620 | 2,594 | -1% | 0 | 0 | — |
case-05 | pass→pass | 8,986 | 6,041 | -33% | 1 | 1 | 0% | 1,513 | 1,409 | -7% | 0 | 0 | — |
case-06 | pass→pass | 13,281 | 6,600 | -50% | 1 | 1 | 0% | 2,164 | 1,667 | -23% | 0 | 0 | — |
case-07 | fail→pass | 15,134 | 8,689 | -43% | 1 | 1 | 0% | 2,448 | 1,946 | -21% | 0 | 0 | — |
case-08 | pass→pass | 13,491 | 12,286 | -9% | 1 | 1 | 0% | 2,263 | 2,502 | +11% | 0 | 0 | — |
case-09 | pass→pass | 11,421 | 9,441 | -17% | 1 | 1 | 0% | 1,906 | 1,933 | +1% | 0 | 0 | — |
case-19 | pass→pass | 7,067 | 4,688 | -34% | 1 | 1 | 0% | 1,511 | 1,276 | -16% | 0 | 0 | — |
case-10 | fail→fail | 7,851 | 5,949 | -24% | 1 | 1 | 0% | 1,556 | 1,413 | -9% | 0 | 0 | — |
case-11 | fail→pass | 8,214 | 5,339 | -35% | 1 | 1 | 0% | 1,403 | 1,353 | -4% | 0 | 0 | — |
case-12 | pass→pass | 8,540 | 5,629 | -34% | 1 | 1 | 0% | 1,784 | 1,382 | -23% | 0 | 0 | — |
case-13 | pass→pass | 8,812 | 7,745 | -12% | 1 | 1 | 0% | 1,739 | 1,869 | +7% | 0 | 0 | — |
case-14 | pass→pass | 5,017 | 1,924 | -62% | 1 | 1 | 0% | 1,062 | 747 | -30% | 0 | 0 | — |
case-15 | pass→pass | 6,136 | 3,329 | -46% | 1 | 1 | 0% | 1,309 | 1,035 | -21% | 0 | 0 | — |
case-16 | pass→pass | 8,560 | 6,914 | -19% | 1 | 1 | 0% | 1,581 | 1,501 | -5% | 0 | 0 | — |
case-17 | pass→pass | 9,923 | 3,194 | -68% | 1 | 1 | 0% | 1,753 | 960 | -45% | 0 | 0 | — |
case-18 | pass→pass | 13,203 | 10,396 | -21% | 1 | 1 | 0% | 2,287 | 1,982 | -13% | 0 | 0 | — |
case-20 | pass→pass | 19,304 | 16,718 | -13% | 1 | 1 | 0% | 4,276 | 4,190 | -2% | 0 | 0 | — |
case-21 | pass→pass | 5,856 | 3,359 | -43% | 1 | 1 | 0% | 1,183 | 1,164 | -2% | 0 | 0 | — |
case-22 | pass→pass | 12,086 | 13,683 | +13% | 1 | 1 | 0% | 2,843 | 3,482 | +22% | 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 +14 percentage points is the difference between those two pass rates over the 22 comparable cases.
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.