agentic-long-run

You orchestrate long-running tasks by executing a milestone pipeline with explicit checkpoints between planning, execution, and review phases. This approach saves persistent state on disk to ensure recoverability and maintain an auditable history across multi-day projects.

Yes, resuming a multi-day workflow after crashes or rate limits is supported natively. The system resumes execution from the last saved checkpoint on disk, bypassing completed milestones to prevent duplicate work and maintain auditable state.

A milestone-driven workflow for multi-day execution structures complex tasks into a sequential pipeline of plan-crafting, run-plan, and review-work stages. It requires persistent state on disk to coordinate dependent milestones and provide recoverable, auditable history.

To start a long run session, initiate the pipeline and approve the initial milestone plans. This begins the coordinated execution flow and establishes the persistent state required for subsequent checkpointing and cross-milestone validation.

Yes, parallel milestone coordination is supported alongside cross-milestone validation. This ensures integration integrity across dependent tasks while maintaining checkpointed progress and resumable execution throughout the multi-day project lifecycle.

The primary limitation for resuming long-running workflows is the requirement for persistent state on disk. Without this environment setup and a suite of agentic skills, recoverability and cross-milestone validation cannot be guaranteed after disruptions.