autonomous-loop

Autonomous story execution is automated by loading stories, dispatching specialist agents, and enforcing the RED → GREEN → REFACTOR cycle. The loop processes stories sequentially, applies retries, and escalates to humans when failures or dependencies block progress.

Autonomous escalation handles failures by writing handoff state and pausing the story loop when dependencies block progress or errors occur. It automatically escalates to humans during assisted or autonomous operation to keep workflows safely on track.

Yes, autonomous story execution requires an approved implementation plan, a stories.yaml file in the workflow context, and workspace readiness. These prerequisites ensure the loop can properly coordinate agent dispatch, testing, and security checks.

Checkpointing during story processing writes progress state to a .checkpoint file after each agent dispatch and quality gate. This mechanism ensures the loop tracks story status, manages retries, and compiles final loop reports for review.

Yes, agent dispatch adapts to both assisted and autonomous modes with upgrade prompts and safety nets. Mode-aware behavior allows the loop to run stories sequentially and auto-escalate when needed, providing flexibility for human-in-the-loop workflows.

You should avoid autonomous story execution if your workspace lacks readiness, an approved implementation plan, or a valid stories.yaml file. The loop requires these elements to properly coordinate TDD enforcement, agent dispatch, and security checks.