What problem does it solve? Coordinating multiple backlog tickets across parallel AI workers is error-prone: dependencies get violated, workers overwrite each other, review feedback goes unanswered, and tracker state drifts from Git reality. This Skill provides a supervised orchestration workflow that dispatches isolated workers per ticket, enforces dependency gates, and keeps humans in control of merges. ## Core Features & Use Cases - Dependency-gated scheduling: Builds the work graph only from explicit blocker references and dispatches a ticket only when every blocker is verifiably merged into the target branch. - Multi-source and multi-harness adapters: Works with GitHub, GitLab, and Jira work sources, and Codex, Grok, or OpenCode worker harnesses, with a bundled Go supervisor for durable unattended runs. - Centralized publication authority: Workers never push, open reviews, or touch the tracker; the supervisor verifies diffs, creates checkpoint commits, pushes via explicit refspecs, and opens draft reviews. - Use Case: Select a Jira Story with Sub-tasks, and the supervisor discovers ready children, launches up to three isolated workers in separate checkouts, publishes draft merge requests as each finishes, processes authorized review feedback, and closes items only after human merges are verified on the target branch. ## Quick Start Ask the agent to implement the ready tickets from the backlog using the implement-tickets orchestration workflow with the appropriate tracker and harness adapters.
