What problem does it solve?
Coordinating multiple independent coding tasks by hand is slow and error-prone: agents duplicate discovery work, branches conflict at merge time, and long-running work loses progress when the session is cleared. This Skill automates the full lifecycle of parallel multi-task work, from planning through merged, tested, documented results.
Core Features & Use Cases
- Planning phase: Gathers code intelligence, analyzes merge-conflict risks, groups tasks into maximally parallel phases, and writes self-contained agent prompts plus PLAN.md and STATE.md files.
- Autonomous execution phase: Dispatches Opus agents in isolated git worktrees, runs a per-task review gauntlet with auto-fix commits, merges each phase behind a test gate, and finishes with a documentation-update agent.
- Crash-safe state management: All state lives in file-backed storage outside the repo, so work resumes cleanly after /clear or session restarts, with hard-exit BLOCKED conditions for human intervention.
- Use Case: You have five independent refactoring and feature tasks for a repo. Run the skill once to generate an approved plan, then let it dispatch agents in parallel worktrees overnight, review and fix each diff, merge phases behind test gates, and update docs.
Quick Start
Ask the assistant to run the parallel-phases skill to plan and execute your list of independent tasks in parallel worktrees with automated review and merging.
