# ralph-loop

> tbd

- Author: Rico Sta. Cruz
- Repository: rstacruz/agentic-toolkit
- Version: 20260119090412
- Stars: 6
- Forks: 0
- Last Updated: 2026-02-06
- Source: https://github.com/rstacruz/agentic-toolkit
- Web: https://mule.run/skillshub/@@rstacruz/agentic-toolkit~ralph-loop:20260119090412

---

---
name: ralph-loop
description: tbd
---

## Ralph loop

A "ralph loop" iterates through a plan in a ticket-by-ticket basis.

Pre-requisites:

1. Ensure there is a PRD as a Markdown file (`artefacts/prd.md` by default)
2. Ensure the PRD has tickets
3. Ensure tickets have acceptance criteria
4. Ensure there is a progress file (`artefacts/progress.md`) - create an empty one if not

Workflow:

1. Spawn a @general agent to do the instructions in `<skill_dir>/once.md`. Use its instructions in verbatim, but change the file paths if needed, and include additional instructions / preamble if given by user.

2. Verify that the agent created a git commit. If it didn't, create one.

3. If the agent reports PRD_COMPLETE, stop. otherwise, repeat step 1 (up to 10 times max).

Guidelines:

- Only 1 iteration per agent. Do not ask agent to do more than 1.