What problem does it solve?
Autonomous development workflows driven by Gemini Code suffer from context fragmentation, fragile CI/merge processes, and coordination complexity when running parallel agents or iterative pipelines; autonomous-loops provides proven patterns and operational guidance to make those loops reliable, resumable, and safe to run at scale.
Core Features & Use Cases
- Sequential pipeline guidance for chaining non-interactive gemini runs to implement, clean up, and verify changes.
- Persistent REPL (NanoClaw) patterns for session-aware exploration and long-lived context.
- Parallel agent orchestration including infinite agentic waves and orchestrator/sub-agent assignment to ensure uniqueness and scalability.
- Continuous PR loop practices for creating branches, auto-fixing CI failures, and automated merge strategies with completion signals and cost/time limits.
- De-sloppify cleanup pass to remove test and code slop as a separate focused step.
- RFC-driven DAG (Ralphinho) for decomposing large specs into dependency DAGs, tiered quality pipelines, eviction-aware merge queues, and resumable worktrees.
- Use Case: Run a continuous-gemini loop that creates iterative PRs, auto-fixes CI failures, and lands incremental improvements with shared task notes bridging iterations.
Quick Start
Ask Gemini to run a sequential pipeline that implements a feature, performs a de-sloppify cleanup pass, runs verification, and creates a conventional commit message.
