SkillFoxx

ralph-starter CLI

CLI that runs autonomous coding loops: a spec from GitHub, Linear or Notion, and an agent working until done.

  • 107 ★ on the repository
  • Last commit
  • MIT
CLI

Broad access

Why this level

  • The agent changes code autonomously in a loop, commits and opens «pull request» while spending tokens.
  • Figma, Notion and Linear need access tokens.
All reasons and checks

rubenmarcus/ralph-starterAuthor:rubenmarcus

Install

In your terminal, with SkillFoxx CLI

npx skillfoxx add cli/ralph-starter

Detects the agents on your machine, shows the access level and pins the version.

Other ways to install

Access and checks

How we review

Broad accessThe tool can change data, run commands or handle secrets.

  • The agent changes code autonomously in a loop, commits and opens «pull request» while spending tokens.
  • Figma, Notion and Linear need access tokens.

This is third-party code. Review the repository files before installing.

What it does

ralph-starter runs a coding agent (Claude Code, Cursor, Copilot, Gemini CLI and others) in a loop and feeds it errors until the task is done. The command takes a spec from GitHub issues, Linear tickets, Notion pages, Figma designs, URLs or local files. Examples: `ralph-starter run --from github --project myorg/myrepo --label "ready"` or `ralph-starter run "build a todo app with React" --commit`. With no arguments a wizard asks about the idea and stack, and `ralph-starter ideas` helps invent a project. There are 16 or more mode presets, including feature, tdd, debug and review. Safeguards include a circuit breaker that stops a stuck loop, a rate limiter and cost tracking. After each iteration tests, lint and build can run, and commits and pull requests are created automatically.

Who it is for. Developers who hand whole tracker tasks to an agent and want a loop with stopping, validation and cost tracking.

Good fit when

  • A working feature has to be built from GitHub issues labeled ready.
  • The spec lives in Notion or Figma and is pulled from there directly.
  • After each agent iteration tests, lint and build should run.
  • A stuck loop should stop on the circuit breaker and token spend should be visible.

Not a fit when

  • The task needs careful manual edits; the loop changes code autonomously.
  • The agent must not be allowed to commit and open pull requests.
  • The token budget is tight and the loop repeats calls until it succeeds.

Example request

Take GitHub issues labeled ready, run the loop in a separate branch with test validation and show the total cost.

Limitations

The loop changes code without intervention and can spend many tokens: watch the cost counter, set limits and work in a branch under version control. The `--commit` and `--pr` options create commits and pull requests on their own, and PRs need `gh`. Different sources need tokens: Figma requires `FIGMA_TOKEN`, Notion needs one for private pages, and GitHub and Linear work through their CLIs or API keys. Node.js 18 or newer, Git and at least one installed coding agent are required. Result quality is set by the agent and the spec, and the loop does not guarantee that the task was done correctly.

How to disable. Remove the CLI with `npm uninstall -g ralph-starter` and disable the Claude Code plugin through `/plugin`. Revoke the Figma and Notion tokens in their settings.

FAQ

How do I install ralph-starter?

Run `npm install -g ralph-starter`. Node.js 18 or newer, Git and at least one coding agent such as Claude Code or Cursor are required.

Which spec sources are supported?

GitHub (through `gh` or a token), Linear, Notion, Figma, any URL with Markdown or HTML and local Markdown and PDF files.

How do I check a source connection?

With `ralph-starter integrations list`, `ralph-starter integrations test github` and `ralph-starter integrations fetch github owner/repo`.

How is a stuck loop stopped?

The circuit breaker stops it after repeated failures. A rate limiter and cost tracking control spend, and `--validate` enables checks after iterations.

What is the Ralph Wiggum technique?

An autonomous coding pattern: the agent runs in a loop, gets errors fed back and retries until the task is complete. Geoffrey Huntley described it.

Discussion

No questions or reviews yet.

CLI tools

A CLI that loops Claude Code, Codex, OpenCode and other agents over a task list or PRD until everything is done

CLIBroad access3K
Needs a VPN

A CLI that takes tickets from GitHub, Linear or Jira, writes code through Claude Code and opens a pull request for your review

CLIBroad access728

Orchestrator for Claude Code, Codex and Gemini CLI: YAML workflows that loop until your own quality gates pass

CLIBroad access334
Official

The @itsaplan/runner CLI takes tasks from the It's a Plan tracker and hands them to Claude Code, Codex or opencode locally.

CLIBroad access2.3K

A background agent for Linear, GitHub, GitLab and Slack tasks; the "/cyrus-setup" skill handles setup.

SkillBroad access895

A Telegram bot drives Claude Code on your computer: you message the chat and the agent edits the project and streams the reply.

CLIBroad access65

README in short

ralph-starter wraps the agent-loop technique in a ready CLI: spec sources, mode presets, safeguards, validation and git automation.

README badge

Report a problem