Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Treat the current session and auto-resume in a new terminal window.
.claude/skills/ruya-ai-reload/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | -53% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -53% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -78% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -43% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -61% | 0% |
Treat the session and spawn an auto-resume watcher that opens a new terminal after you exit.
bash cozempic current --diagnose
bash cozempic treat current -rx $ARGUMENTS If no argument was provided, use standard: bash cozempic treat current -rx standard
bash cozempic reload -rx $ARGUMENTS Do NOT run cozempic treat --execute before cozempic reload — reload already treats internally. Running both would double-treat.
/exit — a new Terminal window will open automatically with the pruned session."| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 6,324 | 4,621 | -27% | 1 | 1 | 0% | 1,100 | 371 | -66% | 0 | 0 | — |
case-02 | fail→fail | 4,500 | 4,144 | -8% | 1 | 1 | 0% | 684 | 353 | -48% | 0 | 0 | — |
case-03 | fail→fail | 5,824 | 6,750 | +16% | 1 | 1 | 0% | 770 | 395 | -49% | 0 | 0 | — |
case-04 | fail→fail | 6,797 | 6,812 | +0% | 1 | 1 | 0% | 1,063 | 568 | -47% | 0 | 0 | — |
case-05 | fail→fail | 7,072 | 3,636 | -49% | 1 | 1 | 0% | 1,003 | 494 | -51% | 0 | 0 | — |
case-06 | fail→fail | 5,539 | 5,590 | +1% | 1 | 1 | 0% | 814 | 393 | -52% | 0 | 0 | — |
case-07 | fail→fail | 5,456 | 2,987 | -45% | 1 | 1 | 0% | 872 | 631 | -28% | 0 | 0 | — |
case-08 | fail→pass | 5,733 | 4,030 | -30% | 1 | 1 | 0% | 1,005 | 472 | -53% | 0 | 0 | — |
case-09 | fail→pass | 6,586 | 2,102 | -68% | 1 | 1 | 0% | 1,256 | 594 | -53% | 0 | 0 | — |
case-10 | fail→pass | 13,529 | 1,787 | -87% | 1 | 1 | 0% | 2,166 | 485 | -78% | 0 | 0 | — |
case-11 | fail→fail | 9,384 | 7,172 | -24% | 1 | 1 | 0% | 1,528 | 716 | -53% | 0 | 0 | — |
case-12 | fail→fail | 21,102 | 5,453 | -74% | 1 | 1 | 0% | 1,746 | 534 | -69% | 0 | 0 | — |
case-13 | fail→pass | 7,604 | 3,651 | -52% | 1 | 1 | 0% | 1,279 | 727 | -43% | 0 | 0 | — |
case-14 | fail→pass | 12,092 | 3,421 | -72% | 1 | 1 | 0% | 2,076 | 819 | -61% | 0 | 0 | — |
case-15 | fail→fail | 18,755 | 3,310 | -82% | 1 | 1 | 0% | 3,175 | 697 | -78% | 0 | 0 | — |
case-16 | fail→pass | 8,399 | 1,062 | -87% | 1 | 1 | 0% | 1,343 | 358 | -73% | 0 | 0 | — |
case-17 | fail→pass | 9,933 | 1,426 | -86% | 1 | 1 | 0% | 1,756 | 401 | -77% | 0 | 0 | — |
case-18 | pass→pass | 9,493 | 3,030 | -68% | 1 | 1 | 0% | 1,466 | 711 | -52% | 0 | 0 | — |
case-19 | fail→pass | 9,465 | 1,370 | -86% | 1 | 1 | 0% | 1,531 | 434 | -72% | 0 | 0 | — |
case-20 | pass→pass | 2,676 | 2,665 | -0% | 1 | 1 | 0% | 453 | 670 | +48% | 0 | 0 | — |
case-21 | pass→pass | 2,490 | 6,173 | +148% | 1 | 1 | 0% | 439 | 665 | +51% | 0 | 0 | — |
case-22 | pass→pass | 2,234 | 8,237 | +269% | 1 | 1 | 0% | 359 | 1,121 | +212% | 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 16 counted toward the lift figure. The other 6 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 +36 percentage points is the difference between those two pass rates over the 16 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.