This walks you through setting up progressive feature rollouts in LaunchDarkly with automated monitoring and rollback triggers. You define stages (typically 1%, 10%, 50%, 100%), pick metrics to watch (error rates, latency, conversions), and set regression thresholds that either notify you or automatically roll back if things go sideways. The workflow is pretty structured: inspect your flag variations, design your stages with monitoring windows, configure metrics with thresholds, then fire it off. It's built for the common case where you want to release gradually but don't want to babysit dashboards 24/7. The automatic rollback on regression detection is the real value here, assuming you trust your thresholds and have metrics that actually reflect user pain.
npx skills add https://github.com/launchdarkly/agent-skills --skill launchdarkly-guarded-rollout