What problem does it solve? Coordinating multiple AI agent sessions manually is chaotic and error-prone. This Skill provides structured patterns for running parallel agent workflows across Claude Code, Codex, OpenCode, and other harnesses using dmux, a tmux-based pane manager. ## Core Features & Use Cases - Parallel Workflow Patterns: Five ready-made patterns including research+implement, multi-file features, test+fix loops, cross-harness work, and parallel code review. - Git Worktree Isolation: Instructions for isolating file changes per agent using git worktrees to avoid merge conflicts. - Worktree Orchestration Helper: A plan.json-driven script that creates branch-backed worktrees, per-worker task files, and tmux panes automatically. - Use Case: Split a large feature into three independent tasks, run each in its own agent pane with its own worktree, then merge the results back into the main branch. ## Quick Start Ask the agent to split the current task into parallel dmux panes, for example by saying "run these three independent tasks in parallel with dmux and merge the results".
