Future ideas and features under consideration.
Agent automatically iterates on CI failures:
- Poll GitHub Actions after each push
- On failure: analyze logs, fix, commit
- Continue until green checkmark
Agent creates/updates issues before coding:
- Problem analysis and proposed approach
- Implementation plan as "thinking out loud" space
- Links PR back to issue for context
Standardized repo setup for mainloop agents:
| Component | Purpose |
|---|---|
| GitHub Actions | CI pipeline (lint, type-check, test) |
| K8s/Helm | Preview environments per PR |
| CNPG operator | Dynamic test databases |
| trunk.yaml | Unified linter config |