Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Drive the canonical spec-kitty next control loop and route query, step, blocked, decision_required, and terminal results.
.claude/skills/priivacy-ai-spk-run-next/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✗→✓ | ▲ Improved | -49% | 0% |
| case-01 | ✗→✓ | ▲ Improved | 3% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -22% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -30% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -59% | 0% |
Use this skill when advancing an active mission, asking what to do next, or recovering from a runtime decision.
spec-kitty next --mission <handle> --json or advance withspec-kitty next --agent <name> --mission <handle> --result <success|failed|blocked>.
query, inspect only; do not execute a prompt or mark a result.step, execute the generated prompt file.decision_required, answer with --answer, --result, --agent, and--decision-id when multiple decisions are pending.
blocked, fix guard failures before retrying.terminal, route to spk-gate-accept.For detailed runtime semantics, use spec-kitty-runtime-next when available.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-06 | fail→pass | 8,760 | 2,497 | -71% | 1 | 1 | 0% | 1,492 | 766 | -49% | 0 | 0 | — |
case-01 | fail→pass | 6,707 | 5,304 | -21% | 1 | 1 | 0% | 1,314 | 1,359 | +3% | 0 | 0 | — |
case-02 | fail→fail | 4,168 | 4,035 | -3% | 1 | 1 | 0% | 733 | 526 | -28% | 0 | 0 | — |
case-03 | fail→fail | 8,236 | 6,316 | -23% | 1 | 1 | 0% | 1,339 | 691 | -48% | 0 | 0 | — |
case-04 | fail→pass | 4,627 | 2,116 | -54% | 1 | 1 | 0% | 741 | 581 | -22% | 0 | 0 | — |
case-05 | fail→pass | 4,486 | 1,931 | -57% | 1 | 1 | 0% | 839 | 585 | -30% | 0 | 0 | — |
case-07 | pass→pass | 9,335 | 2,706 | -71% | 1 | 1 | 0% | 1,783 | 753 | -58% | 0 | 0 | — |
case-08 | pass→pass | 11,897 | 3,209 | -73% | 1 | 1 | 0% | 1,990 | 735 | -63% | 0 | 0 | — |
case-09 | fail→pass | 6,772 | 1,657 | -76% | 1 | 1 | 0% | 1,239 | 505 | -59% | 0 | 0 | — |
case-10 | fail→fail | 12,820 | 1,295 | -90% | 1 | 1 | 0% | 2,173 | 419 | -81% | 0 | 0 | — |
case-11 | fail→pass | 7,766 | 2,606 | -66% | 1 | 1 | 0% | 1,575 | 715 | -55% | 0 | 0 | — |
case-12 | fail→pass | 10,671 | 2,501 | -77% | 1 | 1 | 0% | 2,098 | 731 | -65% | 0 | 0 | — |
case-13 | fail→pass | 12,495 | 2,058 | -84% | 1 | 1 | 0% | 2,406 | 612 | -75% | 0 | 0 | — |
case-14 | pass→pass | 6,611 | 5,018 | -24% | 1 | 1 | 0% | 1,192 | 417 | -65% | 0 | 0 | — |
case-15 | fail→fail | 6,889 | 16,363 | +138% | 1 | 1 | 0% | 1,341 | 1,660 | +24% | 0 | 0 | — |
case-16 | fail→pass | 12,200 | 2,728 | -78% | 1 | 1 | 0% | 2,162 | 706 | -67% | 0 | 0 | — |
case-17 | fail→fail | 9,122 | 8,118 | -11% | 1 | 1 | 0% | 1,693 | 1,778 | +5% | 0 | 0 | — |
case-18 | fail→pass | 7,248 | 3,974 | -45% | 1 | 1 | 0% | 1,144 | 951 | -17% | 0 | 0 | — |
case-19 | pass→pass | 6,885 | 6,719 | -2% | 1 | 1 | 0% | 1,160 | 501 | -57% | 0 | 0 | — |
case-20 | fail→pass | 6,470 | 1,649 | -75% | 1 | 1 | 0% | 1,064 | 507 | -52% | 0 | 0 | — |
case-21 | pass→pass | 9,348 | 2,473 | -74% | 1 | 1 | 0% | 1,593 | 620 | -61% | 0 | 0 | — |
case-22 | fail→pass | 6,170 | 1,592 | -74% | 1 | 1 | 0% | 954 | 435 | -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, and 21 counted toward the lift figure. The other 1 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 +55 percentage points is the difference between those two pass rates over the 21 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.