Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Apply when debugging. Trace each symptom to its root cause and fix it there; reproduce first, ask why until you reach it, resist nil-check guards that silence crashes.
.claude/skills/kunanonj-cursor-plugin-pstack-principle-fix-root-causes/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -10% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -32% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -10% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -2% | 0% |
| case-16 | ✗→✓ | ▲ Improved | -19% | 0% |
When debugging, do not paper over symptoms. Trace every problem to its root cause and fix it there.
Why: Symptom fixes accumulate. Each workaround makes the system harder to reason about, and the real bug remains. Root-cause fixes are slower upfront but reduce total debugging time.
Pattern:
Restart bugs: suspect state before code
Code doesn't change between runs. State does. When something "fails after restart," suspect stale persistent state first: config files, caches, lock files, serialized state. If clearing a state file restores behavior, prioritize state validation as the fix.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 14,725 | 14,495 | -2% | 1 | 1 | 0% | 2,605 | 2,352 | -10% | 0 | 0 | — |
case-02 | pass→pass | 14,438 | 13,859 | -4% | 1 | 1 | 0% | 2,898 | 2,613 | -10% | 0 | 0 | — |
case-03 | fail→pass | 15,588 | 8,963 | -43% | 1 | 1 | 0% | 2,457 | 1,679 | -32% | 0 | 0 | — |
case-04 | fail→pass | 7,396 | 5,444 | -26% | 1 | 1 | 0% | 1,262 | 1,135 | -10% | 0 | 0 | — |
case-05 | fail→fail | 15,258 | 9,755 | -36% | 1 | 1 | 0% | 2,849 | 1,986 | -30% | 0 | 0 | — |
case-06 | pass→pass | 17,787 | 10,147 | -43% | 1 | 1 | 0% | 2,945 | 1,898 | -36% | 0 | 0 | — |
case-07 | pass→pass | 12,625 | 9,233 | -27% | 1 | 1 | 0% | 2,234 | 1,695 | -24% | 0 | 0 | — |
case-08 | pass→pass | 16,967 | 15,505 | -9% | 1 | 1 | 0% | 2,900 | 2,696 | -7% | 0 | 0 | — |
case-09 | fail→pass | 9,984 | 8,871 | -11% | 1 | 1 | 0% | 1,917 | 1,884 | -2% | 0 | 0 | — |
case-10 | pass→fail | 14,012 | 15,053 | +7% | 1 | 1 | 0% | 2,559 | 2,722 | +6% | 0 | 0 | — |
case-16 | fail→pass | 14,673 | 11,138 | -24% | 1 | 1 | 0% | 2,776 | 2,236 | -19% | 0 | 0 | — |
case-11 | pass→pass | 40,545 | 14,833 | -63% | 1 | 1 | 0% | 2,751 | 2,913 | +6% | 0 | 0 | — |
case-12 | pass→pass | 15,253 | 13,604 | -11% | 1 | 1 | 0% | 2,571 | 2,473 | -4% | 0 | 0 | — |
case-13 | fail→fail | 12,431 | 12,001 | -3% | 1 | 1 | 0% | 2,335 | 2,307 | -1% | 0 | 0 | — |
case-14 | pass→pass | 14,173 | 12,462 | -12% | 1 | 1 | 0% | 2,724 | 2,468 | -9% | 0 | 0 | — |
case-15 | pass→fail | 13,477 | 9,786 | -27% | 1 | 1 | 0% | 2,345 | 1,908 | -19% | 0 | 0 | — |
case-17 | fail→pass | 12,043 | 11,157 | -7% | 1 | 1 | 0% | 2,392 | 2,360 | -1% | 0 | 0 | — |
case-18 | pass→pass | 9,216 | 8,572 | -7% | 1 | 1 | 0% | 1,614 | 1,675 | +4% | 0 | 0 | — |
case-19 | fail→pass | 16,277 | 10,673 | -34% | 1 | 1 | 0% | 2,674 | 2,003 | -25% | 0 | 0 | — |
case-20 | pass→pass | 11,680 | 13,975 | +20% | 1 | 1 | 0% | 2,422 | 2,480 | +2% | 0 | 0 | — |
case-21 | pass→pass | 15,776 | 10,812 | -31% | 1 | 1 | 0% | 3,445 | 2,489 | -28% | 0 | 0 | — |
case-22 | pass→pass | 9,168 | 13,288 | +45% | 1 | 1 | 0% | 1,885 | 2,900 | +54% | 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 +23 percentage points is the difference between those two pass rates over the 22 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.