Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Create and manage Inngest functions for reliable background jobs, workflows, and scheduled tasks.
.claude/skills/aiskillstore-inngest-handler/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-09 | ✗→✓ | ▲ Improved | 28% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 29% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 20% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 21% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 46% | 0% |
This skill defines the standards for building durable, multi-step workflows using Inngest.
setTimeout / setInterval:await new Promise(r => setTimeout(r, 1000))await step.sleep("wait-1s", "1s")step.run().let userId; await step.run(..., () => { userId = ... })const userId = await step.run(..., () => { return ... })Wrap all logic in steps to ensure retriability and resumability.
typescriptexport const processOrder = inngest.createFunction( { id: "process-order" }, { event: "shop/order.created" }, async ({ event, step }) => { // 1. Step: Validate (Retriable) const user = await step.run("get-user", async () => { return await db.users.findById(event.data.userId); }); // 2. Step: Sleep (Durable pause) await step.sleep("wait-for-payment", "1h"); // 3. Step: Wait for Event (Human/System interaction) const payment = await step.waitForEvent("wait-payment", { event: "shop/payment.success", match: "data.orderId", timeout: "24h" }); // 4. Step: Conditional Logic if (!payment) { await step.run("cancel-order", async () => { ... }); } } );
Run steps concurrently to speed up execution.
typescriptconst [user, subscription] = await Promise.all([ step.run("fetch-user", () => db.users.find(...)), step.run("fetch-sub", () => stripe.subscriptions.retrieve(...)) ]);
Inside loops, ensure step IDs are unique.
typescriptconst items = event.data.items; for (const item of items) { // Use dynamic ID to ensure uniqueness per item await step.run(`process-item-${item.id}`, async () => { await processItem(item); }); }
Prevent overwhelming 3rd party APIs.
typescriptinngest.createFunction({ id: "sync-crm", // Max 10 requests per minute per user rateLimit: { limit: 10, period: "1m", key: "event.data.userId" }, // Drop events if queue is full throttle: { limit: 5, period: "1s" } }, ...);
Process only the latest event in a window (e.g., search indexing).
typescriptinngest.createFunction({ id: "index-product", // Wait 10s for more events; only run with the latest data debounce: { period: "10s", key: "event.data.productId" } }, ...);
Prioritize specific events (e.g., Paid users).
typescriptinngest.createFunction({ id: "generate-report", // High number = High priority priority: { run: "event.data.plan === 'enterprise' ? 100 : 0" } }, ...);
Inngest retries steps automatically on error (default ~4-5 times with backoff).
{ retries: 10 } in config.Stop execution immediately if the error is fatal (e.g., 400 Bad Request).
typescriptimport { NonRetriableError } from "inngest"; await step.run("validate", async () => { if (!isValid) throw new NonRetriableError("Invalid payload"); });
Execute cleanup logic if the function fails after all retries.
typescriptexport const riskyFunc = inngest.createFunction( { id: "risky-transfer", // Runs if main handler fails onFailure: async ({ error, event, step }) => { await step.run("rollback-funds", async () => { await reverseTransfer(event.data.transferId); }); await step.run("notify-admin", async () => { await sendAlert(`Transfer failed: ${error.message}`); }); } }, { event: "bank/transfer.init" }, async ({ step }) => { /* ... */ } );
MANDATORY: All functions must be imported and exported in src/lib/inngest/functions/index.ts.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-09 | fail→pass | 12,292 | 6,760 | -45% | 1 | 1 | 0% | 2,008 | 2,573 | +28% | 0 | 0 | — |
case-01 | fail→fail | 10,012 | 7,140 | -29% | 1 | 1 | 0% | 2,011 | 2,764 | +37% | 0 | 0 | — |
case-02 | fail→pass | 16,988 | 13,491 | -21% | 1 | 1 | 0% | 3,253 | 4,195 | +29% | 0 | 0 | — |
case-03 | fail→pass | 14,849 | 10,608 | -29% | 1 | 1 | 0% | 3,020 | 3,616 | +20% | 0 | 0 | — |
case-04 | pass→pass | 4,140 | 2,500 | -40% | 1 | 1 | 0% | 728 | 1,808 | +148% | 0 | 0 | — |
case-05 | pass→pass | 4,559 | 3,643 | -20% | 1 | 1 | 0% | 825 | 2,038 | +147% | 0 | 0 | — |
case-06 | fail→pass | 10,423 | 5,601 | -46% | 1 | 1 | 0% | 2,042 | 2,464 | +21% | 0 | 0 | — |
case-07 | pass→pass | 13,590 | 8,483 | -38% | 1 | 1 | 0% | 2,444 | 3,005 | +23% | 0 | 0 | — |
case-08 | fail→pass | 10,885 | 8,600 | -21% | 1 | 1 | 0% | 2,002 | 2,919 | +46% | 0 | 0 | — |
case-10 | fail→pass | 15,665 | 9,213 | -41% | 1 | 1 | 0% | 2,895 | 3,115 | +8% | 0 | 0 | — |
case-11 | fail→pass | 17,724 | 6,297 | -64% | 1 | 1 | 0% | 3,336 | 2,489 | -25% | 0 | 0 | — |
case-12 | fail→pass | 17,136 | 10,792 | -37% | 1 | 1 | 0% | 2,845 | 3,744 | +32% | 0 | 0 | — |
case-13 | fail→pass | 15,906 | 6,942 | -56% | 1 | 1 | 0% | 3,012 | 2,614 | -13% | 0 | 0 | — |
case-14 | pass→pass | 14,422 | 4,557 | -68% | 1 | 1 | 0% | 2,635 | 2,229 | -15% | 0 | 0 | — |
case-15 | pass→pass | 18,671 | 7,493 | -60% | 1 | 1 | 0% | 3,272 | 2,880 | -12% | 0 | 0 | — |
case-16 | fail→pass | 17,312 | 8,944 | -48% | 1 | 1 | 0% | 3,217 | 3,060 | -5% | 0 | 0 | — |
case-17 | fail→pass | 14,815 | 7,902 | -47% | 1 | 1 | 0% | 2,623 | 2,641 | +1% | 0 | 0 | — |
case-18 | fail→pass | 21,455 | 8,307 | -61% | 1 | 1 | 0% | 3,892 | 3,111 | -20% | 0 | 0 | — |
case-19 | pass→pass | 9,770 | 5,867 | -40% | 1 | 1 | 0% | 1,911 | 2,513 | +32% | 0 | 0 | — |
case-20 | fail→pass | 14,368 | 7,584 | -47% | 1 | 1 | 0% | 2,622 | 2,876 | +10% | 0 | 0 | — |
case-21 | pass→pass | 12,553 | 2,888 | -77% | 1 | 1 | 0% | 2,145 | 1,854 | -14% | 0 | 0 | — |
case-22 | fail→pass | 9,827 | 5,810 | -41% | 1 | 1 | 0% | 1,856 | 2,405 | +30% | 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 +64 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.