Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Install the Alchemy SDK and configure API key authentication for Web3 development. Use when setting up blockchain API access, creating an Alchemy app, or configuring multi-chain RPC endpoints. Trigger: "install alchemy", "setup alchemy", "alchemy auth", "alchemy API key".
.claude/skills/jeremylongshore-alchemy-install-auth/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-16 | ✗→✓ | ▲ Improved | 14% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 0% | 0% |
| case-18 | ✗→✓ | ▲ Improved | -13% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -5% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 59% | 0% |
Select and prove the current Alchemy client, key type, network, and authentication boundary before integration work. This workflow produces a reviewable artifact and negative-path evidence before any live side effect.
The legacy alchemy-sdk JavaScript repository is archived and deprecated. New EVM reads use viem or direct documented HTTP APIs; transacting applications and Portfolio support use @alchemy/wallet-apis; Solana work uses Solana Web3.js. An application API key, an Admin access key, a Notify auth token, and a webhook signing key are different credentials and are not interchangeable.
Prefer server-side request-header authentication for confidential workloads. A browser key is acceptable only when the current Alchemy guidance, explicit domain allowlists, least privilege, and the product threat model allow it; short-lived JWTs are the stronger frontend option. Store Admin, Notify, webhook, and signing credentials only in approved secret storage.
viem, direct HTTP, @alchemy/wallet-apis, or Solana Web3.js from the capability—not from the archived SDK examples.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 application owner selects the product and chain. Security approves any browser credential boundary and production secret store. Creating, rotating, revoking, or deleting a live key requires explicit approval from its owner.
Return the capability-to-client decision, chain/feature proof, credential-class matrix, redacted connectivity result, secret lifecycle, owners, evidence date, and unresolved gaps. Mark assumptions, observations, source dates, environment-specific behavior, owners, and unresolved gaps explicitly.
viem, prove one read with a development app key, and keep the production key in the server secret boundary.alchemy-sdk bootstrap even if an old repository still imports it.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-16 | fail→pass | 12,823 | 9,720 | -24% | 1 | 1 | 0% | 2,439 | 2,787 | +14% | 0 | 0 | — |
case-01 | pass→pass | 2,228 | 1,893 | -15% | 1 | 1 | 0% | 396 | 1,293 | +227% | 0 | 0 | — |
case-02 | pass→pass | 5,790 | 1,134 | -80% | 1 | 1 | 0% | 1,065 | 1,168 | +10% | 0 | 0 | — |
case-03 | pass→pass | 4,625 | 4,613 | -0% | 1 | 1 | 0% | 924 | 1,871 | +102% | 0 | 0 | — |
case-04 | fail→pass | 7,402 | 2,371 | -68% | 1 | 1 | 0% | 1,470 | 1,469 | -0% | 0 | 0 | — |
case-05 | pass→pass | 3,676 | 3,797 | +3% | 1 | 1 | 0% | 831 | 1,829 | +120% | 0 | 0 | — |
case-06 | pass→pass | 5,912 | 3,846 | -35% | 1 | 1 | 0% | 1,144 | 1,766 | +54% | 0 | 0 | — |
case-18 | fail→pass | 7,371 | 1,399 | -81% | 1 | 1 | 0% | 1,299 | 1,124 | -13% | 0 | 0 | — |
case-07 | pass→pass | 3,934 | 3,102 | -21% | 1 | 1 | 0% | 823 | 1,662 | +102% | 0 | 0 | — |
case-08 | pass→pass | 2,837 | 2,008 | -29% | 1 | 1 | 0% | 539 | 1,370 | +154% | 0 | 0 | — |
case-09 | pass→pass | 4,202 | 3,296 | -22% | 1 | 1 | 0% | 852 | 1,685 | +98% | 0 | 0 | — |
case-10 | fail→pass | 6,896 | 2,236 | -68% | 1 | 1 | 0% | 1,438 | 1,362 | -5% | 0 | 0 | — |
case-17 | fail→pass | 4,166 | 1,290 | -69% | 1 | 1 | 0% | 733 | 1,168 | +59% | 0 | 0 | — |
case-11 | pass→pass | 2,617 | 1,707 | -35% | 1 | 1 | 0% | 421 | 1,275 | +203% | 0 | 0 | — |
case-12 | pass→fail | 6,145 | 9,575 | +56% | 1 | 1 | 0% | 1,375 | 1,560 | +13% | 0 | 0 | — |
case-13 | pass→pass | 7,226 | 5,547 | -23% | 1 | 1 | 0% | 1,507 | 2,095 | +39% | 0 | 0 | — |
case-14 | pass→pass | 9,286 | 6,021 | -35% | 1 | 1 | 0% | 1,790 | 2,182 | +22% | 0 | 0 | — |
case-15 | pass→pass | 12,656 | 12,563 | -1% | 1 | 1 | 0% | 2,238 | 3,309 | +48% | 0 | 0 | — |
case-19 | pass→pass | 2,655 | 1,248 | -53% | 1 | 1 | 0% | 437 | 1,170 | +168% | 0 | 0 | — |
case-20 | pass→pass | 3,971 | 3,070 | -23% | 1 | 1 | 0% | 817 | 1,670 | +104% | 0 | 0 | — |
case-21 | pass→pass | 4,780 | 3,616 | -24% | 1 | 1 | 0% | 973 | 1,655 | +70% | 0 | 0 | — |
case-22 | pass→pass | 5,772 | 6,986 | +21% | 1 | 1 | 0% | 1,254 | 2,406 | +92% | 0 | 0 | — |
case-23 | pass→pass | 7,129 | 7,675 | +8% | 1 | 1 | 0% | 1,493 | 2,698 | +81% | 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 +17 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.
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.