Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Set up local Web3 development workflow with Alchemy, Hardhat, and testnets. Use when configuring local blockchain dev, testing with Sepolia faucets, or setting up hot-reload for smart contract development. Trigger: "alchemy local dev", "alchemy hardhat", "alchemy testnet setup".
.claude/skills/jeremylongshore-alchemy-local-dev-loop/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -1% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 15% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -8% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 66% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 1% | 0% |
Establish a deterministic local EVM development loop with pinned forks, synthetic fixtures, and bounded Alchemy usage. This workflow produces a reviewable artifact and negative-path evidence before any live side effect.
A remote mainnet fork is reproducible only when the chain, block number, dependency versions, and fixture state are pinned. It still consumes account-level throughput and can diverge when historical access, method support, or upstream state is unavailable. Alchemy Sandbox is a separate simulation option and must be evaluated against the required test semantics.
Use a development-only app key from local secret injection. Never copy a production key or deployer private key into .env, task output, shell history, fixtures, or committed fork URLs.
Use Read, Glob, and Grep to inspect current documentation, configuration, code, fixtures, and evidence. Use Write and Edit only for approved repository artifacts. Skill invocation alone does not authorize network access, credentials, wallet addresses, customer data, plan changes, spend, key creation or rotation, webhook changes, deployment, replay, transaction construction, signing, broadcast, or deletion.
The test owner approves the pinned state and fixture refresh. Security approves secret injection. Forking production state, using customer addresses, or funding a test signer requires data/security approval.
Return the pinned chain/block/dependency matrix, redacted local configuration, deterministic fixtures, positive and clean-room test receipts, usage observation, drift procedure, and rollback. Mark assumptions, observations, source dates, environment-specific behavior, owners, and unresolved gaps explicitly.
Exercise and record expected and observed results for:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-04 | pass→pass | 9,645 | 8,135 | -16% | 1 | 1 | 0% | 2,193 | 2,949 | +34% | 0 | 0 | — |
case-03 | fail→fail | 16,157 | 12,596 | -22% | 1 | 1 | 0% | 3,439 | 3,798 | +10% | 0 | 0 | — |
case-01 | fail→pass | 10,914 | 6,211 | -43% | 1 | 1 | 0% | 2,549 | 2,532 | -1% | 0 | 0 | — |
case-02 | fail→pass | 11,643 | 7,604 | -35% | 1 | 1 | 0% | 2,554 | 2,945 | +15% | 0 | 0 | — |
case-05 | pass→fail | 8,082 | 9,716 | +20% | 1 | 1 | 0% | 1,712 | 3,579 | +109% | 0 | 0 | — |
case-06 | pass→pass | 11,215 | 9,675 | -14% | 1 | 1 | 0% | 2,448 | 3,580 | +46% | 0 | 0 | — |
case-07 | fail→pass | 9,384 | 3,199 | -66% | 1 | 1 | 0% | 1,911 | 1,760 | -8% | 0 | 0 | — |
case-08 | fail→pass | 6,133 | 5,070 | -17% | 1 | 1 | 0% | 1,192 | 1,978 | +66% | 0 | 0 | — |
case-09 | fail→pass | 11,296 | 5,724 | -49% | 1 | 1 | 0% | 2,283 | 2,298 | +1% | 0 | 0 | — |
case-10 | pass→pass | 3,386 | 1,264 | -63% | 1 | 1 | 0% | 642 | 1,374 | +114% | 0 | 0 | — |
case-11 | pass→pass | 4,456 | 1,968 | -56% | 1 | 1 | 0% | 789 | 1,514 | +92% | 0 | 0 | — |
case-12 | fail→pass | 13,181 | 2,683 | -80% | 1 | 1 | 0% | 2,471 | 1,708 | -31% | 0 | 0 | — |
case-13 | fail→pass | 7,552 | 1,929 | -74% | 1 | 1 | 0% | 1,330 | 1,531 | +15% | 0 | 0 | — |
case-14 | fail→pass | 4,818 | 1,972 | -59% | 1 | 1 | 0% | 943 | 1,502 | +59% | 0 | 0 | — |
case-15 | fail→pass | 6,936 | 2,079 | -70% | 1 | 1 | 0% | 1,322 | 1,538 | +16% | 0 | 0 | — |
case-16 | pass→pass | 10,651 | 3,126 | -71% | 1 | 1 | 0% | 2,331 | 1,845 | -21% | 0 | 0 | — |
case-17 | fail→pass | 12,977 | 2,637 | -80% | 1 | 1 | 0% | 2,798 | 1,756 | -37% | 0 | 0 | — |
case-18 | pass→pass | 10,628 | 6,112 | -42% | 1 | 1 | 0% | 2,099 | 2,220 | +6% | 0 | 0 | — |
case-19 | pass→pass | 10,524 | 6,200 | -41% | 1 | 1 | 0% | 1,868 | 2,308 | +24% | 0 | 0 | — |
case-20 | pass→pass | 7,370 | 2,690 | -64% | 1 | 1 | 0% | 1,505 | 1,622 | +8% | 0 | 0 | — |
case-21 | pass→pass | 10,207 | 8,108 | -21% | 1 | 1 | 0% | 1,753 | 2,623 | +50% | 0 | 0 | — |
case-22 | fail→pass | 7,855 | 4,072 | -48% | 1 | 1 | 0% | 1,501 | 1,859 | +24% | 0 | 0 | — |
case-23 | fail→pass | 9,440 | 1,458 | -85% | 1 | 1 | 0% | 1,580 | 1,405 | -11% | 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. The headline lift of +48 percentage points is the difference between those two pass rates over the 23 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
The publisher has shipped newer versions since this run, so these numbers describe v1, not the version currently listed.
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.