linkrunner-sdk-code-auditlisted
Install: claude install-skill linkrunner-labs/skills
# Linkrunner - SDK code audit
Answers **"is Linkrunner wired up correctly in this app?"** by reading the code, whatever the app is
built with. It is read-only: it finds what is wrong, proves it with file:line evidence, and proposes the
fix. To *add* or *rewrite* the integration, use the platform skill (`npx @linkrunner/skills add
<platform>`). To prove behaviour on a device, follow
[SDK Integration Testing](https://docs.linkrunner.io/testing/integration-testing).
## 0. Gather the facts first
Detect or ask for these before judging anything:
- **Which package ships.** In a monorepo, a hit in `examples/` or a legacy app is not the integration.
- **The stack and the pinned SDK version**, from the lockfile rather than the manifest range
(`references/detect-stack.md`). Several methods are version-gated, so "not called" can mean "not
available in your version".
- **The commit or branch** you are auditing. Findings are only true for that tree.
- **Report or fix?** The default is a report. Only edit code when asked.
## 1. Run the scanner
```bash
bash scripts/scan-integration.sh <app-root> # scripts/ is relative to this skill
```
It prints the stack, the SDK dependency and version, every candidate call site with file:line, the
`trackEvent` name inventory, hygiene smells (hardcoded `debug: true`, token literals, unstable
`paymentId`, string `amount`, guarded calls), the platform config, and a per-method coverage count.
**It is a map, not a verdict.** Open and read every lin