What problem does it solve? Coordinating multi-agent software development work—routing tasks to the right specialists, enforcing review and testing gates, and managing commits, merges, and deploys—requires constant manual supervision. This Skill defines a stack-agnostic team-lead orchestrator that runs a fixed 12-stage pipeline autonomously, stopping only at defined human checkpoints. ## Core Features & Use Cases - Fixed 12-stage pipeline: Covers task intake, routing, planning, delegation, code review, documentation, testing, commit, merge, deploy, knowledge revision, and final reporting, with an explicit state-transition table for automatic stage advancement. - Human stop points: Pauses only for plan approval, merge to master, and production deploy; everything else proceeds autonomously without intermediate confirmation questions. - Delegation strategy: Chooses between spawning isolated subagents for large phases and inline skills for trivial work, with subagent nesting limited to one level. - Ralph loop mode: Supports an external restart loop where each iteration runs one phase in a fresh context window, using the todo file's phase table as cross-iteration memory. - Use Case: A project wrapper skill imports this core, fills in its routing table, delegate list, validation commands, and reviewer iteration limit, then runs a fullstack feature from an approved todo file through review, tests, and a single PR merge with only one human stop. ## Quick Start Import this skill from a project wrapper that defines the routing table, delegates, and validation commands, then ask the teamlead to run an approved todo file through the full pipeline.
