Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Run an autonomous /loop iteration -- check progress, work on next task, schedule next wake
.claude/skills/ruvnet-autopilot-loop/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-12 | ✗→✓ | ▲ Improved | -54% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -59% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -51% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -62% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -73% | 0% |
Run one autopilot iteration using Claude Code's native /loop:
mcp__plugin_ruflo-core_ruflo__autopilot_statusmcp__plugin_ruflo-core_ruflo__autopilot_disable and stopmcp__plugin_ruflo-core_ruflo__autopilot_predict for the optimal next actionmcp__plugin_ruflo-core_ruflo__autopilot_logScheduleWakeup({ delaySeconds: 270, reason: "next autopilot iteration" })Always use delay 270s (under 300s cache TTL) to keep the prompt cache warm between iterations.
Autopilot discovers tasks from:
Configure: mcp__plugin_ruflo-core_ruflo__autopilot_config({ taskSources: ["team-tasks", "swarm-tasks"] })
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-12 | fail→pass | 6,959 | 1,830 | -74% | 1 | 1 | 0% | 1,309 | 596 | -54% | 0 | 0 | — |
case-01 | fail→fail | 4,402 | 2,660 | -40% | 1 | 1 | 0% | 265 | 526 | +98% | 0 | 0 | — |
case-02 | fail→fail | 6,004 | 2,780 | -54% | 1 | 1 | 0% | 987 | 568 | -42% | 0 | 0 | — |
case-03 | fail→fail | 6,528 | 2,209 | -66% | 1 | 1 | 0% | 1,153 | 552 | -52% | 0 | 0 | — |
case-04 | fail→pass | 10,428 | 2,405 | -77% | 1 | 1 | 0% | 1,787 | 737 | -59% | 0 | 0 | — |
case-05 | pass→pass | 11,815 | 4,682 | -60% | 1 | 1 | 0% | 1,921 | 1,068 | -44% | 0 | 0 | — |
case-06 | fail→pass | 6,446 | 1,752 | -73% | 1 | 1 | 0% | 1,192 | 590 | -51% | 0 | 0 | — |
case-07 | fail→pass | 7,733 | 1,521 | -80% | 1 | 1 | 0% | 1,384 | 519 | -63% | 0 | 0 | — |
case-08 | fail→pass | 13,757 | 2,060 | -85% | 1 | 1 | 0% | 2,527 | 677 | -73% | 0 | 0 | — |
case-09 | fail→pass | 8,412 | 1,857 | -78% | 1 | 1 | 0% | 1,502 | 623 | -59% | 0 | 0 | — |
case-10 | pass→pass | 9,275 | 1,624 | -82% | 1 | 1 | 0% | 1,692 | 580 | -66% | 0 | 0 | — |
case-11 | pass→pass | 7,365 | 1,470 | -80% | 1 | 1 | 0% | 1,361 | 510 | -63% | 0 | 0 | — |
case-13 | fail→pass | 6,003 | 1,967 | -67% | 1 | 1 | 0% | 1,135 | 708 | -38% | 0 | 0 | — |
case-14 | fail→pass | 5,657 | 2,315 | -59% | 1 | 1 | 0% | 971 | 782 | -19% | 0 | 0 | — |
case-15 | fail→pass | 8,330 | 1,301 | -84% | 1 | 1 | 0% | 1,397 | 508 | -64% | 0 | 0 | — |
case-16 | fail→pass | 7,385 | 7,918 | +7% | 1 | 1 | 0% | 1,314 | 476 | -64% | 0 | 0 | — |
case-17 | pass→pass | 10,281 | 1,812 | -82% | 1 | 1 | 0% | 1,676 | 601 | -64% | 0 | 0 | — |
case-18 | pass→pass | 3,064 | 1,289 | -58% | 1 | 1 | 0% | 462 | 448 | -3% | 0 | 0 | — |
case-19 | fail→pass | 7,075 | 1,697 | -76% | 1 | 1 | 0% | 1,210 | 560 | -54% | 0 | 0 | — |
case-20 | pass→pass | 9,563 | 2,597 | -73% | 1 | 1 | 0% | 1,659 | 703 | -58% | 0 | 0 | — |
case-21 | pass→fail | 7,629 | 4,791 | -37% | 1 | 1 | 0% | 1,408 | 844 | -40% | 0 | 0 | — |
case-22 | pass→pass | 3,074 | 1,419 | -54% | 1 | 1 | 0% | 571 | 551 | -4% | 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 +45 percentage points is the difference between those two pass rates over the 18 comparable cases. 1 case got worse with the skill loaded, and it is 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.