- Home
- Agent skills
- ralph-loop
ralph-loop
Loop Claude on one prompt until a completion promise appears
What it is
Implements the Ralph Wiggum technique: /ralph-loop feeds the same prompt repeatedly via a Stop hook until a stated completion promise is output, so the model iterates on its own earlier work.
Good for
- Iterate until tests pass
- Grind through a well-specified task
- Run unattended refactors
Install or run
/plugin install ralph-loop@claude-plugins-officialCommand as given in the project README. Read the repository before you run anything on your machine.
Worth knowingCan burn many tokens; set an iteration limit and a clear completion promise.
Sources
- https://github.com/anthropics/claude-plugins-official/tree/main/plugins/ralph-loop (project README, read 13 Nov 2026)
- https://api.github.com/repos/anthropics/claude-plugins-official (GitHub API, read 13 Nov 2026)
The summary is ours, written from the README. Spotted an error? See how we work.