pre-agent-task

Advisory safety checks analyze planned agent actions against module permissions using MODULE_MANIFEST.md and DECISION_LOG.md. You run them by invoking the pre-agent-task command, which produces a checklist of conflicts and warnings without blocking execution.

An advisory pre-flight check is a risk-management process that cross-references planned agent operations with permitted actions and prior decision logs. It surfaces conflicts early by generating a checklist, ensuring agent safety without halting the task.

Yes, you can scope safety checks to specific modules by passing directory paths as arguments. This targets the advisory check to relevant module manifests and decision logs, producing a focused checklist for the specified code paths.

No, the advisory pre-flight check does not block execution by default. It enforces advisory gating by documenting detected conflicts and warnings in a checklist, allowing you to review risks before proceeding with the agent task.

You need to cross-check operations against a decision log when planning multi-step agent tasks touching code, infrastructure, databases, or shared resources. This risk-management step surfaces prior warnings and permission conflicts before execution begins.

The primary limitation is that advisory checks are non-blocking. They document conflicts and warnings from module manifests and decision logs, but they do not prevent execution, meaning you must manually review the generated checklist to ensure agent safety.