#7272nd of 73 in Skills, plugins and rules
ralph
Bash loop that reruns Amp or Claude Code until every PRD story passes
- Stars
- 22k
- License
- MIT
- Last commit
- Feb 2026
- Language
- TypeScript
13,429 on 25 Mar 202621,963 on 14 Oct 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
Ralph is a bash script (ralph.sh) that repeatedly launches Amp or Claude Code with a fresh context, each time picking the highest-priority story in prd.json where passes is false. It implements one story, runs typecheck and tests, commits on success, and records learnings in progress.txt and AGENTS.md. It also ships prd and ralph skills for writing a PRD and converting it to prd.json.
Who it is for: Developers who want unattended coding-agent runs against a PRD task list
Strengths
- Fresh context each iteration; state persists only in git, progress.txt and prd.json
- Supports both Amp and Claude Code via a --tool flag
- Small footprint: one shell script, prompt templates and two skills, plus jq
- Installable as a Claude Code plugin marketplace entry
Weaknesses
- Needs working typecheck and tests; broken checks compound across iterations
- Stories must fit one context window; large ones must be split by hand
- Requires an authenticated Amp or Claude Code install, which are separate tools
- No tagged releases; default cap is 10 iterations
What it needs
- no GPU
- Needs Amp CLI or Claude Code, jq, git
- Models: Amp, Claude Code
Also in Skills, plugins and rules
See all 73| Rank | Project | Score |
|---|---|---|
| 1 | skillsAgent skills for planning, TDD, debugging and code review | 94 out of 100 |
| 2 | eccAgents, skills, hooks and rules bundle for Claude Code, Codex and other harnesses | 94 out of 100 |
| 3 | cavemanCuts token use in coding agents with terse replies and compressed inputs | 93 out of 100 |
| #4 | ponytailSkill prompt that makes coding agents write less code and add tests | 92 out of 100 |
| #5 | ui-ux-pro-max-skillDesign-system skill that gives coding agents searchable UI styles, palettes and fonts | 91 out of 100 |