executing-plans

To execute a written implementation plan with verification checkpoints, load the plan into a separate session, review it critically, and run each task step-by-step while pausing on blockers or unclear instructions. The workflow mandates verification before completion and a finishing workflow to safely close the development branch.

The best way to ensure task-by-task progress tracking during software delivery is to execute each planned task as bite-sized steps, mark their status, and run the specific verifications defined by the plan before moving forward. This enforces strict adherence to plan steps and stop-on-failure behavior.

You cannot run your development workflow on the main branch when executing plans. After all tasks pass verification, the workflow requires finishing a development branch using a mandated finishing workflow rather than proceeding on main without consent.

When an AI agent encounters unclear instructions during plan execution, it applies stop-on-failure behavior and pauses to resolve questions before proceeding. This ensures strict adherence to the written plan and prevents deviating from the intended implementation steps.

You need a prewritten plan before starting development workflow execution. The process requires a detailed feature implementation plan that includes verification steps, which is then executed in a separate session with critical review points and verification checkpoints.

Implementation plan execution pauses on blockers to enforce critical review and strict adherence to plan steps. This stop-on-failure behavior ensures unresolved questions or unclear instructions are addressed before proceeding, preventing unchecked errors during software delivery.