What problem does it solve? Manually watching a GitHub backlog for issues ready to be specified, planned, and implemented is repetitive and error-prone. This Skill automates the loop: it polls a repository for issues carrying a chosen label and drives each one through the full speckit flow (specify, clarify, plan, tasks, implement, PR, CI, review) with human approval gates at every phase. ## Core Features & Use Cases - Label-driven issue discovery: Finds open issues tagged with a configurable label, excluding ones already claimed or blocked, and ranks them by age or priority labels. - Three scheduling modes: Run a single sweep (once), an in-session polling loop (timed or self-paced), or a durable cron schedule that re-triggers the skill remotely. - Human-in-the-loop delegation: Delegates each issue to the speckit-gh driver, which posts specs, plans, and task lists on the issue and waits for owner approval before proceeding. - Release issue routing: Detects owner-authorized release issues and routes them to a gitflow release flow instead of feature implementation. - Safety rails: Enforces owner-only directives, caps parallel issues, never auto-merges PRs, and delegates dirty working-tree repair to a fixer subagent. - Use Case: A maintainer labels three backlog issues "ready-for-speckit" and asks the agent to poll every 15 minutes; each issue is specified, planned, implemented, and opened as a PR awaiting their review. ## Quick Start Ask the agent to watch the current GitHub repository for issues labeled ready-for-speckit and run the speckit flow on them one at a time.
