What problem does it solve? Coordinating multiple coding agents on a large project manually is slow and error-prone. This Skill lets you plan a project, dispatch parallel Claude Code agents in isolated git worktrees, monitor their progress, and read structured completion reports through a running Claude DevFleet MCP server. ## Core Features & Use Cases - AI Project Planning: Call plan_project to break a natural-language description into a project with chained missions and dependency DAGs. - Parallel Agent Dispatch: Run up to 3 concurrent agents (configurable via DEVFLEET_MAX_AGENTS), each in an isolated worktree with auto-merge on completion. - Monitoring & Reporting: Track progress with get_mission_status and get_dashboard, then read structured reports (files changed, errors, next steps) via get_report. - Use Case: Ask to "build a REST API with auth and tests" — the Skill plans the missions, dispatches the root mission, auto-dispatches dependent missions as dependencies complete, and reports results back to you. ## Quick Start Connect the DevFleet MCP server with claude mcp add devfleet --transport http http://localhost:18801/mcp, then ask the AI to plan and launch your project description.
