What problem does it solve? Managing a backlog of planned tasks stored as markdown files is error-prone: tasks get skipped, indexes go stale, commits are fragmented, and mistakes repeat. This Skill autonomously executes every pending task in .ai-memory/plans/pending/ in a continuous self-loop until the queue is empty, keeping plans, memory logs, and git history synchronized. ## Core Features & Use Cases - Continuous Multi-Agent Execution: Spawns up to 2 parallel sub-agents with isolated working directories, disjoint file assignments, and a 3-strike rollback rule for failed tasks. - Plan Lifecycle Management: Moves completed plans from pending/ to completed/, consolidates subtasks into single files, and updates all index files in the same turn. - Strict Git & CI/CD Discipline: Enforces atomic grouped commits, immediate pushes, relative-path-only references, and GitMap Pipeline-AI monitoring with dynamic ETA-based waiting instead of rapid polling. - Use Case: A team maintains feature plans as markdown files in .ai-memory/plans/pending/. Invoke this Skill to have the agent read every plan, execute them in sequenced waves (schemas, then services, then UI), record learnings, and push consolidated commits until nothing remains. ## Quick Start Execute all pending tasks in the .ai-memory/plans/pending folder using the execute-pending-tasks workflow and do not stop until the queue is empty.
