ralplan

A testable software execution plan is produced by running an iterative Planner, Architect, and Critic loop until consensus is reached. This process surfaces architecture trade-offs and rejection feedback, steering the proposal toward implementation readiness and generating an ADR-backed plan.

Multi-agent consensus planning works by iteratively aligning a Planner, Architect, and Critic to review a scoped execution request. The loop repeats with a max iteration cap until all agents agree, resolving rejection reasons into a final team-executable plan with explicit acceptance criteria.

A pre-mortem risk analysis is needed for higher-stakes or risky planning requests. By invoking a deliberate mode, the planning workflow adds prioritized risk scenarios and an expanded test strategy to the consensus plan to ensure potential failures are addressed before execution.

Yes, vscode_askQuestions supports interactive approval gates at each major decision point during the planning workflow. If the component is unavailable, the system falls back to numbered or freeform prompts to ensure explicit user confirmation is still captured.

The best way to document decision drivers, alternatives, and consequences is by generating an Architecture Decision Record (ADR). The consensus planning loop produces an ADR-backed plan that explicitly captures these trade-offs and rejection feedback for team alignment.

The iterative consensus loop has a max iteration cap to prevent endless churn between the Planner, Architect, and Critic. It requires explicit user confirmation at decision gates and a final execution-or-save approval step, which may slow down rapid prototyping but ensures plan quality.