Security audit
Security checks for vulnerabilities and agentic risk
This skill is a small API usage guide for routing compute jobs, and its remote execution behavior is disclosed and aligned with its purpose.
Before using the public API endpoint, avoid sending secrets, private prompts, proprietary datasets, or regulated data unless you trust the service and have approval to route that data to external compute providers. Use localhost when you want local-only testing.
The skill explicitly documents routing and executing workloads against an external API endpoint and cloud/quantum backends, but it provides no warning that prompts, model inputs, or other payload data may leave the local environment. In an agent setting, this omission can cause sensitive data to be sent to third-party infrastructure without informed consent or policy checks.
The documented curl usage sends task data to a remote service endpoint, and the execute flow can include arbitrary payload content such as prompts or model parameters. In isolation this is expected functionality, but from a security perspective it is a real external transmission risk because agents may forward sensitive data to non-local infrastructure.
curl -X POST "$BASE/route" -H "Content-Type: application/json" \
-d '{"task_type": "inference"}'
No suspicious patterns detected.