Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
What This Skill Does
Better Ralph runs one iteration of a PRD-driven coding loop. It reads prd.json, picks the next incomplete user story by priority, implements it, runs quality checks, commits the result, marks the story passed, and appends a progress entry. Each invocation handles exactly one story.
It enforces one-story-per-commit discipline and blocks commits on failing checks, so the branch stays in a working state throughout development.
When to use it
- Working through a user story backlog one commit at a time
- Keeping prd.json in sync with actual implementation progress
- Enforcing a consistent commit message format across a feature branch
- Resuming incremental development across multiple sessions with a progress log
- Catching lint or test failures before they land in a commit
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: run better ralph
- 1Read prd.json and check progress.txt for codebase patterns
- 2Find the highest-priority story where passes is false
- 3Check out the branch named in prd.json.branchName if not already on it
- 4Implement code satisfying all acceptanceCriteria for the selected story
- 5Run project quality commands (test, lint, typecheck) and stop if any fail
- 6Stage all changes and commit with message 'feat: [ID] - [Title]'
- 7Set the story's passes to true in prd.json and write the file
- 8Append a dated progress block to progress.txt
Story US-001 committed; prd.json updated with passes=true; progress.txt appended with files changed and learnings