ralph_loop
Start an iterative AI development loop that continuously improves work through self-referential feedback, with progress tracking, git integration, and template support.
Instructions
Start a Ralph Wiggum iterative development loop.
Ralph is a development methodology based on continuous AI agent loops. The technique creates a self-referential feedback loop where the same prompt is fed back repeatedly, allowing the AI to iteratively improve its work until completion.
NEW FEATURES:
Iteration history tracking with progress metrics
Git integration for automatic change tracking
External tool integration (tests, linters)
Smart stagnation detection and warnings
Pre-built templates for common tasks
Use ralph_list_templates to see available templates.
Input Schema
TableJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
| prompt | No | The task prompt to iterate on (can be omitted if using template_id) | |
| template_id | No | ID of a pre-built template to use (overrides prompt) | |
| max_iterations | No | Maximum iterations before auto-stop (0 = unlimited) | |
| completion_promise | No | Promise phrase that signals completion (e.g., 'DONE', 'COMPLETE'). When detected in output as <promise>PROMISE</promise>, the loop ends. | |
| git_enabled | No | Enable git integration (default: true) | |
| auto_commit | No | Automatically commit changes after each iteration (default: false) |