executing-plans

Phase-by-phase plan execution prevents drift by enforcing mandatory verification gates and confidence scoring. It reads the plan source of truth, blocks completion until goal evidence passes, and refuses to continue if scope boundaries or acceptance criteria are violated.

Plan confidence scoring is a per-phase requirement that gates software delivery by blocking progression until verification checks pass. It updates task ledgers with receipts and ensures required goals have explicit evidence or a user-approved waiver before completion.

You can block completion without evidence by maintaining goal and task ledgers that require receipts. The execution stops if goal evidence, verifier ownership, and drift guard checks are missing, preventing the AI session from finishing work without proof.

Yes, you can use scope-safety and verification gates for AI coding sessions. The execution refuses to start or continue when missing acceptance criteria, scope boundaries, or plan source drift would make results unreliable, ensuring strict adherence to the reviewed plan.

You should not use automated plan execution when your plan lacks Goal IDs, dependencies, scoped verification commands, or explicit stop conditions. Without these required inputs, the confidence gates and drift guard checks cannot reliably verify phase completion.

Yes, executing-plans requires a pre-approved implementation plan source of truth. It reads this reviewed plan to enforce phase-by-phase execution, applying to AI-assisted software delivery only when tasks include explicit Goal IDs and scoped verification commands.