sqs-testslisted
Install: claude install-skill testland/qa
# sqs-tests
## Overview
[sqs-dg]: https://docs.aws.amazon.com/AWSSimpleQueueService/latest/SQSDeveloperGuide/welcome.html
Two queue types, per [the SQS developer guide][sqs-dg]:
| Type | Delivery semantics |
|---|---|
| **Standard** | "at-least-once message delivery" |
| **FIFO** | "exactly-once message processing" + "high-throughput" mode |
The semantic difference cascades through every test pattern: a
Standard-queue test must assume duplicates can occur; a FIFO test
must assume strict ordering.
## When to use
- The repo has SQS-publishing or SQS-consuming code (AWS SDK).
- The user writes unit tests against mocked SQS clients.
- The user writes integration tests against LocalStack / real SQS
(in a sandbox account).
- A test verifies visibility-timeout, DLQ-routing, or
long-polling behavior.
## Step 1 - Test approach: mock vs LocalStack vs real
Three approaches, ordered by isolation:
| Approach | Pros | Cons |
|---|---|---|
| `aws-sdk-client-mock` (TS) / `moto` (Python) | Pure unit, no network | Doesn't catch AWS-side behavior (visibility timeouts, DLQ routing) |
| LocalStack (Docker SQS emulator) | Full SQS semantics locally | Slower; not 100% behavior parity with AWS SQS |
| Real SQS in sandbox AWS account | Highest fidelity | Costs money; per-PR queue cleanup needed |
For pure logic tests (does the code call SendMessage with the right
body?), use mocks. For semantic tests (does retry-on-failure work
end-to-end?), use LocalStack. For pre-prod smoke, use real