executing-plans

Execute approved plan steps exactly as written by processing one step at a time and logging progress after each. This prevents scope creep by raising notes for related improvements instead of bundling them into the current execution.

Plan execution pauses for user input if a step fails, reveals a mismatch, or requires a decision. This stop condition prevents proceeding with incorrect assumptions and ensures the approved workflow remains intact.

No, plan execution enforces one-step-at-a-time processing without bundling steps. Each step runs independently to maintain strict adherence to the approved plan and log accurate progress after every step.

Guardrails are needed when executing an approved plan with numbered steps to prevent scope expansion mid-execution. They enforce exact execution by stopping for user input on mismatches or decisions rather than auto-expanding the scope.

Yes, the execution applies when a user has a plan with numbered steps and explicit approval to proceed, including mid-execution contexts. It resumes by running the next pending step exactly as written without expanding scope.

The best way to manage an approved plan is executing steps exactly as written, one at a time, while logging progress after each step. This ensures no steps are skipped and scope is not expanded beyond what was approved.