How we built a software factory to drive Astro’s GitHub issue count to zero

Astro’s maintainers built an automated GitHub Actions pipeline that cut open issues from over 200 to about 30—and expect to reach zero—by having isolated AI subagents independently reproduce bugs, diagnose root causes, and ship verified fixes as preview releases. The approach did not simply close stale tickets; instead, failure cases exposed gaps in test coverage, missing code comments, or unclear abstractions, which the team fixed to improve both the bot and the codebase for human contributors. The resulting framework, Flue, and the standalone `triagebot-action` are released openly as a reference for other projects wanting to apply the same state‑machine, label‑driven automation to their own issue backlogs without the maintainer burnout that now‑rampant AI‑generated tickets can cause.