Skip to content

Observability And Security

The runtime is where execution-layer security and observability become concrete.

That means two things happen in the same place:

  • the system enforces boundaries around what a task is allowed to do
  • the system emits the evidence operators need to understand what it did

Runtime protections include:

  • workspace boundary enforcement
  • shell validation
  • secret redaction
  • scoped credential injection
  • resource-limited container execution

These are execution concerns. They are different from control-plane auth and workflow policy.

The runtime exposes and emits:

  • health endpoints
  • metrics
  • tracing
  • structured execution logs
  • result and artifact metadata

Operators need to know both whether the system is healthy and what happened during a specific execution. The runtime produces that evidence. The platform stores and presents it.

Runtime observability and security only become useful when they are joined back to operator surfaces:

  • Live Logs and Diagnostics present much of the emitted execution evidence to operators
  • Workflow Detail ties that evidence back to one workflow, task graph, or intervention path
  • Operate Security Model explains the broader platform-plus-runtime security posture beyond execution-layer safeguards alone

The runtime produces the raw evidence and enforces the execution boundary. The platform turns that into durable, searchable, operator visible state.