plan-exec-loop

To execute AI deployment plans safely, you can use a gap-driven plan-exec loop that maintains per-phase sessions and applies strict git branch checks before edits. This orchestrates multi-phase execution while ensuring proposed commits are governed by source.

Session-based resumption works by running dedicated control sessions and per-phase execution sessions with checkpoint persistence. The gap-driven engine resumes the AI planning workflow exactly where the previous session left off, preventing duplicate task execution.

Yes, you can draft database migrations to SCRIPTS.sql for read-only validation. The workflow enforces strict non-execution of DDL and DML statements, ensuring that drafted migrations are only used for safe, read-only analyses during the execution loop.

The best way to manage artifact promotion is through explicit export steps. Artifacts remain securely within sessions during the plan execution loop, and you must use explicit export-* actions to promote them to the docs directory, ensuring zero automatic exports.

Yes, the plan-exec loop approach supports robust git policy enforcement during automated workflows. It performs branch checks before allowing edits and groups proposed commits by their source, maintaining an auditable and safe deployment workflow.

You can resume an interrupted multi-phase AI workflow by leveraging the plan-exec loop's checkpoint persistence. It uses gap-driven execution to identify incomplete phases and resumes the session-based workflow without losing prior progress or repeating completed tasks.