Before execution
Noxgild validates the target computer, capability schema, worker availability, risk metadata, and policy requirements before work is dispatched.
SECURITY
Computer access designed around user control.
Noxgild combines local execution, bounded capabilities, explicit approvals, human verification boundaries, and durable command state.
Security is part of the execution model.
Approved computer identity
Connected computers use account-bound device identity so work targets an explicitly authorized machine.
Bounded capability catalog
Requests map to a curated catalog with defined schemas, worker routing, risk, retry, and verification behavior.
Per-command approval
Sensitive or destructive commands can remain queued until that exact command is explicitly approved.
Human security challenges
CAPTCHA, OTP/2FA, passkeys, secure desktop, and similar protected verification stay human.
Local execution
The local runtime performs file, process, browser, and desktop work on the authorized computer.
Scoped authentication
Noxgild uses authenticated account and device flows, short-lived connection grants, and scoped access tokens.
Untrusted-content boundary
Browser-derived content cannot silently pivot into Machine or Desktop control without additional approval.
Execution uncertainty is explicit
If Noxgild cannot prove whether a side effect finished, the command can become UNKNOWN rather than being guessed or replayed.
During execution
The computer independently checks catalog security metadata and policy before the local worker handles a command.
After execution
The command record keeps an authoritative result or an explicit uncertain state so clients can verify before retrying consequential work.
SECURITY REPORTING
Found a security issue?
Please report security concerns privately to support@noxgild.com. Noxgild also publishes a standard security.txt file for automated discovery.