What problem does it solve? Coordinating parallel AI coding workers on shared repositories is error-prone: workers edit the wrong checkout, report under stale tokens, or merge without verification. This Skill turns a Codex session into a disciplined worker that executes one authorized assignment, reports evidence, and waits for integration approval. ## Core Features & Use Cases - Assignment intake and validation: Reads a Gru assignment packet, verifies the worktree, ticket, token, and done-when criteria, and acknowledges with an exact completion contract before starting work. - Structured progress reporting: Sends ack, progress, blocked, ready, and complete reports over Hermod peer messaging with the engagement, task, and token identity. - Verified delivery workflow: Runs required checks and code review, posts findings, and waits for an integration grant before merging and closing the ticket. - Use Case: A team lead dispatches three parallel feature tasks to separate Codex sessions; each Minion works in its own worktree, rebases onto the moving base, and reports a reviewed draft PR for the coordinator to merge. ## Quick Start Invoke the minion skill with a Gru assignment packet containing the engagement, task, token, worktree, objective, and done-when criteria to begin the assigned implementation.