What problem does it solve? AI coding sessions often stall by asking for confirmation between tasks, skip verification, or report completion prematurely. This Skill runs an approved implementation plan from the first task to the last without interruption, verifying each step and reporting only once at the end. ## Core Features & Use Cases - Autonomous task loop: Executes every task in an approved plan.md (T1 through Tn) without asking for re-confirmation, resuming from any specified task such as "continue from T6". - Per-task verification: Runs build, test, and static checks from AGENTS.md after each task, including red-state confirmation for new test cases, plus a final full verification and completion-reviewer subagent review. - Defined stop conditions and reporting: Halts only for destructive operations, repeated failures, secrets in diffs, or out-of-plan scope changes, then produces a structured completion report with actual verification output. - Use Case: After approving a plan to add a password-reset feature across five files, tell the agent to proceed; it implements each task, commits per task, runs the full test suite, and delivers one final report with HUMAN-VERIFY items flagged. ## Quick Start Approve the proposed plan and tell the agent to proceed with the implementation to run all tasks through final verification.
