ralph-opencode-free-loop

What This Skill Does

The ralph-opencode-free-loop skill provides an automated, autonomous coding environment within OpenClaw. It leverages the Ralph Wiggum CLI tool to execute complex programming tasks while utilizing OpenCode Zen as the primary agent provider. The skill is specifically designed to handle multi-step development workflows, such as bug fixing, refactoring, or feature implementation, by maintaining an iterative loop until specific success criteria are met or a completion promise is signaled. A standout feature of this skill is its robust, built-in fallback mechanism; it automatically cycles through a curated list of free OpenCode models (kimi-k2.5-free, minimax-m2.1-free, glm-4.7-free, and big-pickle) to ensure that your development process is never interrupted by transient rate limits, model unavailability, or quota issues.

Installation

To install this skill, use the ClawHub CLI tool in your terminal: clawhub install openclaw/skills/skills/bderiel/open-ralph. Ensure you have the ralph CLI installed globally in your environment and that your current working directory is a initialized Git repository, as the tool requires version control for tracking autonomous changes.

Use Cases

This skill is ideal for developers who require a hands-off approach to repetitive or complex coding tasks. It is best used for:

• Automated Bug Resolution: Automatically iterating through failed test suites until all tests return a passing state.
• Refactoring Workflows: Applying consistent code cleanup across large repositories without manual oversight.
• Error Squashing: Running a loop that targets specific linter or type-checker errors until the build succeeds.
• Multi-Iteration Debugging: Allowing the agent to attempt different logic paths when encountering tricky environmental or logic-based bugs.

Example Prompts

1. "Ralph, please refactor the authentication middleware to use the new session context and ensure all unit tests pass. Use the standard iteration loop."
2. "Fix all TypeScript compilation errors in the /src directory. Stop only when tsc reports zero errors and the build command exits with code 0."
3. "Add a new integration test for the login flow and then update the existing handler to support the new payload structure. Complete the loop when the build succeeds."

Tips & Limitations

• Safety First: Always review code changes generated by the agent. The tool performs file-writes, so ensure you commit or stage your work before running a long loop.
• Verifiable Criteria: Always define specific, verifiable success criteria within your prompt (e.g., 'tests pass', 'tsc passes') to prevent infinite loops.
• Fallback Awareness: If the primary model fails due to 'quota exceeded', the system will automatically try the next model in the priority chain. Use the --no-plugins flag if you experience unexpected behavior or conflicts with existing OpenCode configurations.
• Git Dependency: This skill requires a Git repository to manage state and roll back changes if an autonomous operation goes wrong.