Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Orchestrate end-to-end backend feature development from requirements to deployment. Use when coordinating multi-phase feature delivery across teams and services.
.claude/skills/dokhacgiakhoa-backend-development-feature-development/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -20% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -41% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -20% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 23% | 0% |
| case-20 | ✗→✓ | ▲ Improved | 110% | 0% |
Orchestrate end-to-end feature development from requirements to production deployment:
Extended thinking: This workflow orchestrates specialized agents through comprehensive feature development phases - from discovery and planning through implementation, testing, and deployment. Each phase builds on previous outputs, ensuring coherent feature delivery. The workflow supports multiple development methodologies (traditional, TDD/BDD, DDD), feature complexity levels, and modern deployment strategies including feature flags, gradual rollouts, and observability-first development. Agents receive detailed context from previous phases to maintain consistency and quality throughout the development lifecycle.]
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 36,371 | 42,213 | +16% | 1 | 1 | 0% | 8,266 | 6,609 | -20% | 0 | 0 | — |
case-02 | pass→pass | 16,698 | 7,949 | -52% | 1 | 1 | 0% | 2,792 | 2,154 | -23% | 0 | 0 | — |
case-03 | pass→pass | 12,888 | 9,756 | -24% | 1 | 1 | 0% | 2,281 | 2,141 | -6% | 0 | 0 | — |
case-04 | pass→pass | 10,020 | 11,832 | +18% | 1 | 1 | 0% | 2,011 | 2,985 | +48% | 0 | 0 | — |
case-05 | fail→fail | 46,283 | 32,398 | -30% | 1 | 1 | 0% | 8,233 | 6,408 | -22% | 0 | 0 | — |
case-06 | pass→pass | 41,622 | 22,774 | -45% | 1 | 1 | 0% | 8,237 | 4,596 | -44% | 0 | 0 | — |
case-07 | fail→fail | 38,776 | 29,216 | -25% | 1 | 1 | 0% | 7,562 | 6,145 | -19% | 0 | 0 | — |
case-08 | pass→fail | 20,033 | 19,959 | -0% | 1 | 1 | 0% | 3,204 | 3,995 | +25% | 0 | 0 | — |
case-09 | fail→fail | 32,961 | 24,648 | -25% | 1 | 1 | 0% | 5,278 | 4,506 | -15% | 0 | 0 | — |
case-10 | pass→fail | 41,579 | 38,568 | -7% | 1 | 1 | 0% | 7,367 | 6,254 | -15% | 0 | 0 | — |
case-11 | fail→pass | 42,214 | 30,251 | -28% | 1 | 1 | 0% | 8,219 | 4,830 | -41% | 0 | 0 | — |
case-12 | pass→fail | 38,181 | 27,156 | -29% | 1 | 1 | 0% | 6,428 | 4,276 | -33% | 0 | 0 | — |
case-13 | fail→fail | 25,432 | 29,180 | +15% | 1 | 1 | 0% | 5,059 | 5,381 | +6% | 0 | 0 | — |
case-14 | fail→fail | 23,697 | 31,639 | +34% | 1 | 1 | 0% | 4,306 | 6,004 | +39% | 0 | 0 | — |
case-15 | fail→pass | 41,560 | 29,852 | -28% | 1 | 1 | 0% | 8,212 | 6,568 | -20% | 0 | 0 | — |
case-16 | fail→pass | 34,819 | 39,559 | +14% | 1 | 1 | 0% | 5,886 | 7,259 | +23% | 0 | 0 | — |
case-17 | fail→fail | 24,451 | 22,211 | -9% | 1 | 1 | 0% | 4,183 | 3,925 | -6% | 0 | 0 | — |
case-18 | fail→fail | 24,615 | 24,059 | -2% | 1 | 1 | 0% | 4,260 | 4,400 | +3% | 0 | 0 | — |
case-19 | fail→fail | 33,677 | 50,245 | +49% | 1 | 1 | 0% | 7,427 | 6,579 | -11% | 0 | 0 | — |
case-20 | fail→pass | 25,267 | 44,255 | +75% | 1 | 1 | 0% | 3,584 | 7,515 | +110% | 0 | 0 | — |
case-21 | fail→fail | 36,932 | 30,978 | -16% | 1 | 1 | 0% | 5,662 | 6,741 | +19% | 0 | 0 | — |
case-22 | fail→fail | 26,556 | 39,081 | +47% | 1 | 1 | 0% | 4,229 | 6,391 | +51% | 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 +9 percentage points is the difference between those two pass rates over the 22 comparable cases. 4 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.