Security: harden agent provisioning command execution

[sign-mark]

Summary

Agent provisioning constructs a shell command from request-derived values before executing the spin-up script. This permits shell interpretation of untrusted input and can expose provisioning secrets through captured script output.

Fix

• Execute the configured script with an argument array rather than a shell command string.
• Reject unsafe identifiers used in generated endpoint filenames.
• Fail closed when required provisioning configuration is absent.
• Enforce execution timeout and propagate script failures.
• Avoid logging captured script output that may contain secrets.

Validation

• Added regression tests for argument-array execution, unsafe identifier rejection, and script failure propagation.
• agent-provisioning build passes.