Every 30 minutes, an autonomous agent we call Animus wakes up and sweeps all eight LaunchApp template repositories. It checks health gates, syncs dependency updates, fixes build failures, and opens PRs — all without human intervention.
Here's how it works.
The conductor loop
The conductor runs on a cron schedule. Every 30 minutes it:
- Pulls the latest state of all repos
- Runs build, lint, and test checks against each
- Prioritizes failures by severity (build failures first, then security, then test regressions)
- Dispatches fix tasks to worker agents
Fixing build failures
When a build fails, the conductor identifies the root cause — usually a TypeScript error or a breaking API change in a dependency. It creates a worktree, fixes the issue, runs the health checks again, and opens a PR if the fix passes.
Syncing features
New features land in the flagship React Router template first. The conductor then diffs the flagship against each variant (Next.js, Nuxt, SvelteKit) and opens sync PRs for any missing packages or patterns. A human reviews and merges — the hard work of finding what's missing is automated.
Security scanning
Every sweep runs a dependency audit. Known CVEs are flagged, prioritized by severity, and a fix task is queued. High and critical vulnerabilities get immediate attention.
What Animus doesn't do
Animus doesn't make product decisions. It doesn't design new features. It doesn't write marketing copy. It keeps the trains running so the team can focus on building.