Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Launch both thermo-nuclear review subagents in parallel, then synthesize their findings. Use for thermos, double thermo review, or combined bug/security and code-quality branch audits.
.claude/skills/kunanonj-cursor-plugin-thermos-thermos/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | -38% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -53% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -37% | 0% |
| case-22 | ✗→✓ | ▲ Improved | -50% | 0% |
| case-17 | ✓→✗ | ▼ Worse | -52% | 0% |
Run the two thermo review passes as async background subagents in parallel, then synthesize their results.
run_in_background: true:subagent_type: "thermo-nuclear-review-subagent" for bugs, breakages, security, devex regressions, feature-flag leaks, and other branch-audit risks.subagent_type: "thermo-nuclear-code-quality-review-subagent" for maintainability, structure, file-size growth, spaghetti, abstractions, and codebase-health risks.If individual background summaries are already visible to the user, do not restate them wholesale. Surface the unified verdict, the highest-signal findings, and any remaining uncertainty.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-17 | pass→fail | 9,237 | 2,590 | -72% | 1 | 1 | 0% | 1,338 | 639 | -52% | 0 | 0 | — |
case-01 | fail→fail | 14,199 | 16,772 | +18% | 1 | 1 | 0% | 1,104 | 1,764 | +60% | 0 | 0 | — |
case-02 | fail→fail | 4,511 | 16,171 | +258% | 1 | 1 | 0% | 712 | 481 | -32% | 0 | 0 | — |
case-03 | fail→fail | 7,401 | 4,524 | -39% | 1 | 1 | 0% | 419 | 524 | +25% | 0 | 0 | — |
case-04 | pass→pass | 7,193 | 7,954 | +11% | 1 | 1 | 0% | 1,381 | 1,756 | +27% | 0 | 0 | — |
case-05 | fail→fail | 2,622 | 8,656 | +230% | 1 | 1 | 0% | 389 | 1,176 | +202% | 0 | 0 | — |
case-06 | pass→fail | 11,932 | 31,735 | +166% | 1 | 1 | 0% | 2,447 | 5,067 | +107% | 0 | 0 | — |
case-07 | pass→pass | 9,842 | 2,645 | -73% | 1 | 1 | 0% | 1,674 | 782 | -53% | 0 | 0 | — |
case-08 | fail→pass | 12,199 | 7,359 | -40% | 1 | 1 | 0% | 1,095 | 675 | -38% | 0 | 0 | — |
case-09 | fail→pass | 7,888 | 2,391 | -70% | 1 | 1 | 0% | 1,290 | 609 | -53% | 0 | 0 | — |
case-10 | fail→fail | 12,367 | 3,028 | -76% | 1 | 1 | 0% | 1,939 | 817 | -58% | 0 | 0 | — |
case-11 | pass→pass | 5,802 | 1,982 | -66% | 1 | 1 | 0% | 875 | 612 | -30% | 0 | 0 | — |
case-12 | fail→pass | 12,661 | 6,460 | -49% | 1 | 1 | 0% | 2,192 | 1,389 | -37% | 0 | 0 | — |
case-13 | pass→pass | 9,455 | 6,670 | -29% | 1 | 1 | 0% | 1,531 | 1,121 | -27% | 0 | 0 | — |
case-14 | pass→pass | 9,837 | 4,291 | -56% | 1 | 1 | 0% | 1,606 | 1,005 | -37% | 0 | 0 | — |
case-15 | pass→pass | 12,447 | 6,907 | -45% | 1 | 1 | 0% | 2,025 | 1,275 | -37% | 0 | 0 | — |
case-16 | pass→pass | 8,782 | 2,713 | -69% | 1 | 1 | 0% | 1,463 | 737 | -50% | 0 | 0 | — |
case-18 | pass→fail | 7,579 | 2,011 | -73% | 1 | 1 | 0% | 1,156 | 611 | -47% | 0 | 0 | — |
case-19 | pass→pass | 12,359 | 7,651 | -38% | 1 | 1 | 0% | 2,153 | 1,527 | -29% | 0 | 0 | — |
case-20 | pass→pass | 13,097 | 6,756 | -48% | 1 | 1 | 0% | 2,055 | 1,314 | -36% | 0 | 0 | — |
case-21 | pass→pass | 13,013 | 5,031 | -61% | 1 | 1 | 0% | 2,002 | 1,046 | -48% | 0 | 0 | — |
case-22 | fail→pass | 7,496 | 1,670 | -78% | 1 | 1 | 0% | 1,145 | 572 | -50% | 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 18 counted toward the lift figure. The other 4 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 18 comparable cases. 3 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.