ralph-json-start-loop

Automating feature delivery from PRD JSONs is handled by running the Ralph autonomous loop, which isolates execution across git worktrees, tracks progress, and generates automated commits and PRs for each completed story.

Dependency-aware scheduling in iterative execution works by respecting interdependent story requirements within the PRD JSON, ensuring that prerequisite stories are completed before dependent tasks are executed across isolated worktrees.

Yes, you need jq and gh installed along with git and claude, as these dependencies are required to parse PRD JSON configurations, manage worktree logs, and automate PR generation and status synchronization.

The best way to limit iterations for controlled experimentation is to specify iteration limits when starting the Ralph loop, allowing you to test autonomous execution and PRD state updates on a specific project without running indefinitely.

PRD state needs updating after each iteration to track progress accurately across interdependent stories, ensuring the autonomous loop knows which features are complete and can synchronize status before generating automated commits and PRs.