Coding Agent Loops
by Chelsea AI
Description
Coding Agent Loops
Run AI coding agents in persistent, self-healing sessions with automatic retry and completion notification.
Core Concept
Instead of one long agent session that stalls or dies, run many short sessions in a loop. Each iteration starts fresh — no accumulated context. The agent picks up where it left off via files and git history.
What You Get
- Persistent tmux sessions that survive disconnects and restarts
- Ralph retry loops that automatically restart failed agents
- PRD-based workflows with checklist-driven completion tracking
- Completion hooks that notify you when work is done
- Parallel agent support for independent tasks
Why This Matters
Every AI coding agent eventually stalls — rate limits, token errors, context overflow. This skill turns fragile one-shot runs into robust production workflows. Set it up, walk away, come back to finished code.
Works With
- Codex CLI (OpenAI)
- Claude Code (Anthropic)
- Any agent that can run in a terminal
Battle-tested across hundreds of multi-step builds. Free forever.
What's Included
- SKILL.md — Complete skill definition with commands and patterns
- Quick start templates for common workflows
- PRD format guide for checklist-driven completion
- Troubleshooting guide
Requirements
- •tmux installed
- •ralphy-cli (npm install -g ralphy-cli)
- •A coding agent (Codex CLI or Claude Code)
- •OpenClaw runtime (for completion hooks)
Reviews
“Chelsea has completely transformed how I handle leads. I used to lose half my prospects to slow follow-up.”
“The CMA creator alone saved me 3 hours per listing presentation. Worth every penny.”
“Great foundation. My team of 5 agents all use it. The pipeline manager is a game-changer.”
one-time purchase
Most agents see ROI in the first week
Chelsea AI
We build AI assistants for real estate professionals and operators. Founded on 10+ years of real estate media experience, now bringing AI-powered operations to every agent.