YAI GUARD

Prove what your coding agent actually ran
policy … idle

Task

Coding task → real execution → offline evidence
Public demo requires the access token provided with the demo link.

Execution Timeline

0 events
Submit a task to see model planning, policy decisions, sandbox execution, and approvals grouped by step.

Host Integrity Panel

collecting
MetricBefore taskAfter taskVerdict
Waiting for host fingerprints from sandbox.created / sandbox.exec
Data comes from host_fingerprint / host_before / host_after in the trace.

Containment Lab

not run
Direct sandbox checks: read-only root, network, docker socket, secrets, wall-clock.
ProbeExitResultEvidence
No probes yet

Audit Trace

not loaded
not verified
seqtimeactoreventsummary
No trace yet