ralph-evolver

What This Skill Does

Ralph-Evolver is an advanced recursive self-improvement engine designed to act as an automated architect for your codebase. Unlike standard linting tools or static analysis scripts, it operates on a philosophy of first-principles thinking. It ingests multi-dimensional context, including git commit history, legacy TODO/FIXME markers, and error-handling patterns, to identify the root causes of architectural decay rather than simply applying surface-level patches. By utilizing a recursive loop, the agent can critique its own previous improvements, ensuring that the system evolves toward greater maintainability and structural integrity over time. It continuously asks fundamental questions—such as what a project is missing or how it could be rebuilt from scratch—to force emergence-based design shifts.

Installation

To integrate this agent skill into your OpenClaw environment, execute the following command in your terminal: clawhub install openclaw/skills/skills/hsssgdtc/ralph-evolver Ensure you have the necessary runtime permissions configured for the agent to access your project directories and perform file operations.

Use Cases

• Legacy Refactoring: When dealing with undocumented or "spaghetti" code, Ralph-Evolver helps identify which sections are most prone to failure and suggests deep structural refactors.
• Architectural Debt Reduction: Use the tool to systematically clear out accumulation of technical debt, using the "evolution-level" tracker to ensure you are fixing root design issues.
• Self-Optimizing Workflows: By running the evolver on its own configuration files, you create a meta-loop where the agent becomes more effective at diagnosing your specific codebase patterns the longer it operates.
• Clean-Slate Prototyping: Leverage the "first principles" mode to generate hypothetical designs for complex modules, helping you rethink inefficient features.

Example Prompts

1. "Ralph, analyze the current authentication module. Compare its design against my repository's history and suggest an evolution-level rewrite that prioritizes security and modularity."
2. "Perform a 3-cycle recursive analysis on the /src directory. Identify the top three hotspots contributing to runtime exceptions and propose a refactoring plan that resolves the underlying design patterns."
3. "Evaluate the last 5 improvements made by your engine. Are we drifting toward surface-level fixes or evolution-level changes? Adjust your focus accordingly."

Tips & Limitations

• Safety: Because this tool performs automated code modifications, always run it within a feature branch or on a project with a solid git backup.
• Recursion Depth: Setting the --loop count too high can consume significant LLM credits. Start with a loop of 1 or 2 to verify the agent's direction before scaling up.
• Contextual Noise: The engine works best in repositories with clean commit messages. If your history is messy, it may struggle to extract accurate "why" signals for your current codebase structure.