01
It starts by itself
new PRD file added. Runs in Claude Code routine.
- ↻ new PRD file added
event: new PRD markdown in the ralph directory
Every plan becomes small, ordered tasks.
Install the Ralph skill and make it a habit: follow https://gethabit.dev/skills/snarktank/ralph.md
876 installs22K starsMIT licenceSource on GitHub ↗

Ralph
by snarktank · 876 installs
The skill
Turns a plan document into small steps an agent can run.
Only when you ask.
Ship faster
Every plan becomes small, ordered tasks. Less rework and fewer stuck runs.
↻ Runs by itselfevery time a new plan is added
When does it start? How does it check its work? What does it remember?
Stops after one PRD per run, with at most one retry.
If prd.json exists with a different branchName and progress.txt has content, copy both into archive/YYYY-MM-DD-feature-name/ and reset progress.txt.
Cut the PRD into stories small enough for one Ralph iteration.
Put schema first, then backend, then UI, then summary views.
Give each story checkable criteria, add Typecheck passes, and add the dev-browser skill check for UI stories.
Write prd.json with sequential ids, priorities and empty notes.
Confirm no story depends on a later one and no criterion is vague.
It writes only in its own folder and asks you before anything goes out.
Read both lists. If the left one sounds like you, the Ralph skill helps.
Ralph starts a fresh Amp instance per iteration with no memory, so a big story can run out of context.
No. It only converts the PRD; the ralph.sh script runs the stories.
Copy the prompt at the top into your agent. It installs the skill from snarktank/ralph and sets up the habit. Or run npx skills add snarktank/ralph --skill ralph.
Yes. It is published under MIT. Habit is free too.
Install the Ralph skill and make it a habit: follow https://gethabit.dev/skills/snarktank/ralph.md
All snarktank skills →More planning skills →
Works in
Made by snarktank, published under MIT. Habit links to the original and never changes it. Is this your skill?