What problem does it solve? Large development tasks often stall because they require manual breakdown, coordination of multiple workstreams, and constant supervision. This Skill automates the full lifecycle: capturing the user's request, decomposing it into subtasks, dispatching parallel execution agents, and looping until completion. ## Core Features & Use Cases - Two-Phase Autonomous Loop: Planning mode captures the prompt, scans the codebase with subagents, and decomposes work into subtasks; Execution mode dispatches parallel agents and iterates until done. - Failure Memory & Rollback: Crashes are logged to .ai-memory/memory/issues/ with rollback, so repeated mistakes are avoided across runs. - CI/CD Pipeline Monitoring: Integrates with GitMap Pipeline-AI (gitmap pipeline-ai status) using adaptive ETA-based wait intervals instead of busy-polling. - Use Case: Hand the agent a large feature request like "add authentication to the API"; it writes a plan file, splits it into subtasks, runs two execution agents in parallel, tracks changed files, monitors the remote pipeline, and produces a single atomic commit. ## Quick Start Ask the agent to autonomously plan and execute your full feature request using the parent task N-step loop until all subtasks are complete.
