Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Execute an approved Maestro implementation plan using the shared session-state contract
.claude/skills/josstei-execute/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | -77% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -51% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -60% | 0% |
| case-19 | ✗→✓ | ▲ Improved | 38% | 0% |
| case-23 | ✗→✓ | ▲ Improved | -73% | 0% |
Read ../../references/runtime-guide.md. Call get_skill_content with resources: "execution", "delegation", "session-management", "validation"].
get_runtime_context appears in your available tools, call it first.docs/maestro.docs/maestro as the workspace state root.Read the approved implementation plan at the user-provided path (or check docs/maestro/plans/ for the most recent plan). Resolve the execution mode gate, create or resume session state, then execute phases through child agents following the loaded methodology.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-20 | fail→fail | 9,452 | 3,201 | -66% | 1 | 1 | 0% | 1,400 | 324 | -77% | 0 | 0 | — |
case-01 | fail→fail | 4,331 | 4,986 | +15% | 1 | 1 | 0% | 271 | 612 | +126% | 0 | 0 | — |
case-02 | fail→fail | 4,118 | 2,342 | -43% | 1 | 1 | 0% | 194 | 354 | +82% | 0 | 0 | — |
case-03 | fail→fail | 5,793 | 3,700 | -36% | 1 | 1 | 0% | 296 | 222 | -25% | 0 | 0 | — |
case-04 | fail→fail | 2,774 | 3,868 | +39% | 1 | 1 | 0% | 233 | 207 | -11% | 0 | 0 | — |
case-05 | fail→pass | 17,787 | 5,375 | -70% | 1 | 1 | 0% | 2,712 | 637 | -77% | 0 | 0 | — |
case-06 | fail→pass | 10,338 | 3,928 | -62% | 1 | 1 | 0% | 1,742 | 852 | -51% | 0 | 0 | — |
case-07 | pass→fail | 2,191 | 4,024 | +84% | 1 | 1 | 0% | 276 | 382 | +38% | 0 | 0 | — |
case-08 | fail→pass | 7,368 | 2,170 | -71% | 1 | 1 | 0% | 1,133 | 450 | -60% | 0 | 0 | — |
case-09 | pass→fail | 13,923 | 4,174 | -70% | 1 | 1 | 0% | 2,067 | 451 | -78% | 0 | 0 | — |
case-21 | pass→pass | 4,399 | 3,680 | -16% | 1 | 1 | 0% | 587 | 759 | +29% | 0 | 0 | — |
case-10 | pass→fail | 3,731 | 5,074 | +36% | 1 | 1 | 0% | 558 | 996 | +78% | 0 | 0 | — |
case-11 | pass→pass | 4,084 | 5,804 | +42% | 1 | 1 | 0% | 546 | 608 | +11% | 0 | 0 | — |
case-12 | fail→fail | 8,062 | 5,650 | -30% | 1 | 1 | 0% | 1,202 | 426 | -65% | 0 | 0 | — |
case-13 | fail→fail | 19,182 | 4,552 | -76% | 1 | 1 | 0% | 1,019 | 371 | -64% | 0 | 0 | — |
case-14 | pass→fail | 22,856 | 3,015 | -87% | 1 | 1 | 0% | 4,070 | 222 | -95% | 0 | 0 | — |
case-15 | fail→fail | 7,305 | 11,012 | +51% | 1 | 1 | 0% | 864 | 1,790 | +107% | 0 | 0 | — |
case-16 | pass→fail | 15,724 | 6,351 | -60% | 1 | 1 | 0% | 3,075 | 323 | -89% | 0 | 0 | — |
case-17 | fail→fail | 7,935 | 8,852 | +12% | 1 | 1 | 0% | 1,264 | 1,599 | +27% | 0 | 0 | — |
case-18 | fail→fail | 16,300 | 5,278 | -68% | 1 | 1 | 0% | 2,506 | 440 | -82% | 0 | 0 | — |
case-19 | fail→pass | 3,642 | 6,207 | +70% | 1 | 1 | 0% | 530 | 734 | +38% | 0 | 0 | — |
case-22 | fail→fail | 10,278 | 4,976 | -52% | 1 | 1 | 0% | 1,534 | 385 | -75% | 0 | 0 | — |
case-23 | fail→pass | 9,348 | 3,790 | -59% | 1 | 1 | 0% | 1,337 | 356 | -73% | 0 | 0 | — |
case-24 | fail→fail | 12,802 | 5,791 | -55% | 1 | 1 | 0% | 1,784 | 424 | -76% | 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. 24 cases were attempted, and 10 counted toward the lift figure. The other 14 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 0 percentage points is the difference between those two pass rates over the 10 comparable cases. 5 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.