Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Expert backend architect specializing in scalable API design, microservices architecture, and distributed systems.
.claude/skills/davila7-backend-architect/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 60% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 60% | 0% |
| case-22 | ✓→✗ | ▼ Worse | 204% | 0% |
| case-06 | ✓→✓ | ＝ Same ✓ | 130% | 0% |
| case-07 | ✓→✓ | ＝ Same ✓ | 122% | 0% |
You are a backend system architect specializing in scalable, resilient, and maintainable backend systems and APIs.
Expert backend architect with comprehensive knowledge of modern API design, microservices patterns, distributed systems, and event-driven architectures. Masters service boundary definition, inter-service communication, resilience patterns, and observability. Specializes in designing backend systems that are performant, maintainable, and scalable from day one.
Design backend systems with clear boundaries, well-defined contracts, and resilience patterns built in from the start. Focus on practical implementation, favor simplicity over complexity, and build systems that are observable, testable, and maintainable.
When designing architecture, provide:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-05 | fail→fail | 25,663 | 65,352 | +155% | 1 | 1 | 0% | 4,252 | 7,828 | +84% | 0 | 0 | — |
case-01 | fail→pass | 37,581 | 35,181 | -6% | 1 | 1 | 0% | 6,240 | 10,003 | +60% | 0 | 0 | — |
case-02 | fail→pass | 54,765 | 34,603 | -37% | 1 | 1 | 0% | 6,228 | 9,967 | +60% | 0 | 0 | — |
case-03 | fail→fail | 35,966 | 34,168 | -5% | 1 | 1 | 0% | 6,208 | 9,981 | +61% | 0 | 0 | — |
case-04 | fail→fail | 20,267 | 22,492 | +11% | 1 | 1 | 0% | 3,662 | 7,935 | +117% | 0 | 0 | — |
case-06 | pass→pass | 22,487 | 42,444 | +89% | 1 | 1 | 0% | 3,679 | 8,466 | +130% | 0 | 0 | — |
case-07 | pass→pass | 22,648 | 22,405 | -1% | 1 | 1 | 0% | 3,319 | 7,377 | +122% | 0 | 0 | — |
case-08 | pass→pass | 19,834 | 26,614 | +34% | 1 | 1 | 0% | 3,104 | 7,094 | +129% | 0 | 0 | — |
case-09 | pass→pass | 17,294 | 32,497 | +88% | 1 | 1 | 0% | 2,735 | 7,122 | +160% | 0 | 0 | — |
case-10 | pass→pass | 22,334 | 41,597 | +86% | 1 | 1 | 0% | 3,456 | 7,452 | +116% | 0 | 0 | — |
case-16 | pass→pass | 18,083 | 25,125 | +39% | 1 | 1 | 0% | 2,934 | 7,972 | +172% | 0 | 0 | — |
case-11 | pass→pass | 22,147 | 25,010 | +13% | 1 | 1 | 0% | 3,740 | 8,070 | +116% | 0 | 0 | — |
case-12 | pass→pass | 23,075 | 27,120 | +18% | 1 | 1 | 0% | 2,430 | 8,351 | +244% | 0 | 0 | — |
case-13 | pass→pass | 49,996 | 33,830 | -32% | 1 | 1 | 0% | 5,156 | 9,159 | +78% | 0 | 0 | — |
case-14 | pass→pass | 20,402 | 25,002 | +23% | 1 | 1 | 0% | 3,304 | 7,798 | +136% | 0 | 0 | — |
case-15 | pass→pass | 52,088 | 22,993 | -56% | 1 | 1 | 0% | 4,194 | 7,813 | +86% | 0 | 0 | — |
case-17 | fail→fail | 25,998 | 51,413 | +98% | 1 | 1 | 0% | 4,246 | 8,555 | +101% | 0 | 0 | — |
case-18 | pass→pass | 18,716 | 21,955 | +17% | 1 | 1 | 0% | 2,975 | 7,233 | +143% | 0 | 0 | — |
case-19 | pass→pass | 17,428 | 29,088 | +67% | 1 | 1 | 0% | 2,919 | 8,550 | +193% | 0 | 0 | — |
case-20 | pass→pass | 8,587 | 12,684 | +48% | 1 | 1 | 0% | 1,582 | 5,711 | +261% | 0 | 0 | — |
case-21 | pass→pass | 5,946 | 6,664 | +12% | 1 | 1 | 0% | 1,072 | 4,883 | +356% | 0 | 0 | — |
case-22 | pass→fail | 8,038 | 6,056 | -25% | 1 | 1 | 0% | 1,581 | 4,803 | +204% | 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. The headline lift of +5 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.