Security & Trust
How a run is isolated, what happens to your credentials, and what's audited.
Testing a live target — your own, or a repository you've connected — means Vulnix necessarily handles sensitive material: access credentials, source code, and the traffic the agent generates. Here's what actually happens to it.
Run isolation
Every run executes in its own dedicated sandbox, torn down when the run ends:
Prop
Type
Orphan cleanup
If a run's sandbox is ever left behind by an interrupted workflow, a reconciliation process detects and tears it down automatically — nothing is left running silently.
Credentials and secrets
Authentication
- Client accounts: standard session-based auth, with role-based access per organization (see Roles & Access).
- Staff access: a fully separate portal, gated by mandatory two-factor authentication with no grace period — staff cannot sign in without it, ever.
- Every staff action is audit-logged — who, what, and when.
What's not built yet
We'd rather tell you plainly than let you assume: there is no independent, hardware-level virtualization boundary around a run's sandbox today — isolation is enforced at the container, network, and credential-scope level described above, not a separate virtual machine per run. If that boundary matters to your specific threat model, talk to us before running Vulnix against a target where it's a hard requirement.
