Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Design technical architecture, select technology stack, and define implementation strategy from specifications and constitution constraints.
.claude/skills/a5c-ai-planning-design/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -27% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -26% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -52% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -72% | 0% |
| case-23 | ✗→✓ | ▲ Improved | -74% | 0% |
Architecture decisions must be traceable to specification requirements. Technology choices must comply with constitution constraints. Trade-offs must be documented for every significant decision.
Invoke via babysitter process: methodologies/spec-kit/spec-kit-planning Full pipeline: methodologies/spec-kit/spec-kit-orchestrator
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 22,018 | 14,787 | -33% | 1 | 1 | 0% | 4,068 | 2,960 | -27% | 0 | 0 | — |
case-02 | fail→fail | 26,635 | 23,576 | -11% | 1 | 1 | 0% | 4,728 | 4,356 | -8% | 0 | 0 | — |
case-03 | fail→fail | 22,220 | 19,517 | -12% | 1 | 1 | 0% | 3,747 | 3,451 | -8% | 0 | 0 | — |
case-04 | fail→fail | 16,313 | 16,117 | -1% | 1 | 1 | 0% | 3,357 | 3,235 | -4% | 0 | 0 | — |
case-05 | fail→fail | 18,035 | 17,186 | -5% | 1 | 1 | 0% | 4,174 | 3,529 | -15% | 0 | 0 | — |
case-06 | fail→fail | 12,096 | 10,594 | -12% | 1 | 1 | 0% | 2,313 | 2,211 | -4% | 0 | 0 | — |
case-07 | fail→fail | 17,722 | 19,552 | +10% | 1 | 1 | 0% | 3,031 | 3,358 | +11% | 0 | 0 | — |
case-08 | fail→fail | 21,328 | 18,772 | -12% | 1 | 1 | 0% | 3,522 | 3,765 | +7% | 0 | 0 | — |
case-09 | fail→pass | 24,540 | 16,914 | -31% | 1 | 1 | 0% | 4,448 | 3,277 | -26% | 0 | 0 | — |
case-10 | fail→pass | 19,011 | 1,885 | -90% | 1 | 1 | 0% | 994 | 482 | -52% | 0 | 0 | — |
case-11 | fail→pass | 10,789 | 1,642 | -85% | 1 | 1 | 0% | 1,636 | 451 | -72% | 0 | 0 | — |
case-12 | fail→fail | 20,828 | 16,074 | -23% | 1 | 1 | 0% | 3,712 | 3,054 | -18% | 0 | 0 | — |
case-13 | fail→fail | 17,058 | 17,554 | +3% | 1 | 1 | 0% | 2,990 | 3,290 | +10% | 0 | 0 | — |
case-14 | fail→fail | 21,486 | 17,999 | -16% | 1 | 1 | 0% | 3,700 | 3,556 | -4% | 0 | 0 | — |
case-15 | fail→fail | 18,645 | 14,335 | -23% | 1 | 1 | 0% | 3,189 | 2,605 | -18% | 0 | 0 | — |
case-16 | fail→fail | 17,192 | 18,106 | +5% | 1 | 1 | 0% | 3,108 | 3,228 | +4% | 0 | 0 | — |
case-17 | fail→fail | 19,604 | 16,573 | -15% | 1 | 1 | 0% | 3,544 | 2,807 | -21% | 0 | 0 | — |
case-18 | fail→fail | 21,182 | 13,414 | -37% | 1 | 1 | 0% | 3,347 | 2,674 | -20% | 0 | 0 | — |
case-19 | fail→fail | 32,876 | 29,543 | -10% | 1 | 1 | 0% | 3,214 | 4,549 | +42% | 0 | 0 | — |
case-20 | fail→fail | 25,053 | 29,902 | +19% | 1 | 1 | 0% | 4,559 | 6,343 | +39% | 0 | 0 | — |
case-21 | fail→fail | 19,713 | 20,144 | +2% | 1 | 1 | 0% | 3,187 | 3,708 | +16% | 0 | 0 | — |
case-22 | fail→fail | 24,717 | 21,600 | -13% | 1 | 1 | 0% | 4,158 | 3,953 | -5% | 0 | 0 | — |
case-23 | fail→pass | 12,441 | 2,616 | -79% | 1 | 1 | 0% | 2,349 | 603 | -74% | 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. 23 cases were attempted, and 22 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 +22 percentage points is the difference between those two pass rates over the 22 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.