Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Build production-ready Web3 applications, smart contracts, and decentralized systems. Implements DeFi protocols, NFT platforms, DAOs, and enterprise blockchain integrations. Use PROACTIVELY for smart contracts, Web3 apps, DeFi protocols, or blockchain infrastructure.
.claude/skills/dokhacgiakhoa-blockchain-developer/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-11 | ✗→✓ | ▲ Improved | 54% | 0% |
| case-08 | ✓→✓ | ＝ Same ✓ | 26% | 0% |
| case-18 | ✓→✓ | ＝ Same ✓ | 28% | 0% |
| case-02 | ✓→✓ | ＝ Same ✓ | 60% | 0% |
| case-03 | ✓→✓ | ＝ Same ✓ | -43% | 0% |
resources/implementation-playbook.md.You are a blockchain developer specializing in production-grade Web3 applications, smart contract development, and decentralized system architectures.
Expert blockchain developer specializing in smart contract development, DeFi protocols, and Web3 application architectures. Masters both traditional blockchain patterns and cutting-edge decentralized technologies, with deep knowledge of multiple blockchain ecosystems, security best practices, and enterprise blockchain integration patterns.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-08 | pass→pass | 15,852 | 17,004 | +7% | 1 | 1 | 0% | 2,664 | 3,364 | +26% | 0 | 0 | — |
case-18 | pass→pass | 10,776 | 8,731 | -19% | 1 | 1 | 0% | 1,714 | 2,189 | +28% | 0 | 0 | — |
case-01 | fail→fail | 19,676 | 40,516 | +106% | 1 | 1 | 0% | 3,169 | 4,538 | +43% | 0 | 0 | — |
case-02 | pass→pass | 27,147 | 35,173 | +30% | 1 | 1 | 0% | 4,789 | 7,664 | +60% | 0 | 0 | — |
case-03 | pass→pass | 41,219 | 19,442 | -53% | 1 | 1 | 0% | 6,526 | 3,690 | -43% | 0 | 0 | — |
case-04 | fail→fail | 16,961 | 10,332 | -39% | 1 | 1 | 0% | 3,039 | 2,450 | -19% | 0 | 0 | — |
case-05 | fail→fail | 14,889 | 15,071 | +1% | 1 | 1 | 0% | 3,028 | 3,815 | +26% | 0 | 0 | — |
case-06 | fail→fail | 18,205 | 15,889 | -13% | 1 | 1 | 0% | 3,263 | 2,975 | -9% | 0 | 0 | — |
case-07 | pass→pass | 10,239 | 8,979 | -12% | 1 | 1 | 0% | 2,039 | 2,213 | +9% | 0 | 0 | — |
case-09 | pass→pass | 12,121 | 10,254 | -15% | 1 | 1 | 0% | 1,862 | 2,426 | +30% | 0 | 0 | — |
case-10 | pass→pass | 8,547 | 9,197 | +8% | 1 | 1 | 0% | 1,471 | 1,972 | +34% | 0 | 0 | — |
case-11 | fail→pass | 15,975 | 19,829 | +24% | 1 | 1 | 0% | 2,586 | 3,978 | +54% | 0 | 0 | — |
case-12 | pass→pass | 12,429 | 11,232 | -10% | 1 | 1 | 0% | 2,456 | 2,625 | +7% | 0 | 0 | — |
case-13 | pass→pass | 9,961 | 8,037 | -19% | 1 | 1 | 0% | 1,397 | 1,922 | +38% | 0 | 0 | — |
case-14 | pass→pass | 7,542 | 9,493 | +26% | 1 | 1 | 0% | 1,352 | 2,047 | +51% | 0 | 0 | — |
case-15 | pass→pass | 10,533 | 8,237 | -22% | 1 | 1 | 0% | 1,654 | 1,948 | +18% | 0 | 0 | — |
case-16 | pass→pass | 8,904 | 8,784 | -1% | 1 | 1 | 0% | 1,613 | 1,937 | +20% | 0 | 0 | — |
case-17 | pass→pass | 13,778 | 12,268 | -11% | 1 | 1 | 0% | 2,379 | 2,561 | +8% | 0 | 0 | — |
case-19 | pass→pass | 16,167 | 12,274 | -24% | 1 | 1 | 0% | 2,180 | 2,569 | +18% | 0 | 0 | — |
case-20 | pass→pass | 15,011 | 15,159 | +1% | 1 | 1 | 0% | 2,345 | 2,858 | +22% | 0 | 0 | — |
case-21 | pass→pass | 8,613 | 8,207 | -5% | 1 | 1 | 0% | 1,199 | 1,813 | +51% | 0 | 0 | — |
case-22 | pass→pass | 15,331 | 15,402 | +0% | 1 | 1 | 0% | 2,745 | 2,936 | +7% | 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.
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.