← Back to Skills Marketplace
82
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install ralph-wiggum-loop
Description
Iteracyjnie doskonali kod lub tekst AI, wykrywając i naprawiając błędy, optymalizacje, bezpieczeństwo i styl w maksymalnie trzech krokach.
Usage Guidance
This skill appears to implement the advertised iterative improvement loop, but review and test before using on sensitive code. Specific recommendations:
- Ensure LM Studio runs locally and do NOT set LMSTUDIO_URL to a remote host unless you intend to send your code there (the scripts will transmit the full code to whatever LMSTUDIO_URL is used).
- Install required tools first: Python 3.9+, the 'requests' package (pip install requests), curl, and jq (the shell scripts use jq but SKILL.md didn't mention it).
- Be aware of a small implementation mismatch: ralph-loop.sh passes flags (-u, -s) that generator.py's CLI doesn't define; run the Python modules directly or inspect/fix the shell script before relying on it.
- Review the scripts' behavior (especially fix_code which posts code and issues to the LM endpoint) on non-sensitive examples to confirm behavior and outputs.
- If you plan to run this in a production environment, run it in an isolated environment and audit network traffic to confirm LM Studio is local.
Capability Analysis
Type: OpenClaw Skill
Name: ralph-wiggum-loop
Version: 1.0.0
The skill bundle implements a legitimate iterative code improvement loop (Generator-Critic-Fixer-Verifier) designed to work with a local LM Studio instance. The scripts (ralph-loop.sh, critic.py, and generator.py) perform standard API interactions and include basic security linting to detect SQL injection and hardcoded secrets in the generated code. No evidence of data exfiltration, malicious execution, or harmful prompt injection was found; the preference for 'uncensored' models is contextually aligned with providing a 'ruthless' code critic.
Capability Assessment
Purpose & Capability
Name/description (iterative code/text improvement) matches the included components (generator.py, critic.py, ralph-loop.sh) and the declared runtime behavior (send code to LLM, get issues, fix, verify). No unrelated credentials or services are requested.
Instruction Scope
Runtime instructions and scripts read user-supplied code/files and send them to an LLM endpoint (LM Studio). That is expected for this skill. However: (1) the SKILL.md and scripts assume LM Studio runs at http://127.0.0.1:1234 but the code honors LMSTUDIO_URL/--api-url overrides — if LMSTUDIO_URL is pointed to a remote host, user code will be transmitted off-host; (2) the scripts call external tools (curl, jq) and Python packages (requests) but the SKILL.md omits jq and the Python dependency; (3) there are CLI argument mismatches between ralph-loop.sh and generator.py (the shell passes -u/-s which generator.py's argparse does not define), which is an implementation inconsistency that can cause failures.
Install Mechanism
No install spec is provided (instruction-only deployment). Included files are local scripts and Python modules; there are no downloads from arbitrary URLs or archive extraction. Risk from install mechanism is low, but running the code writes nothing special to disk beyond user-specified outputs.
Credentials
The skill requests no secrets and declares no required env vars. In practice the code uses LMSTUDIO_URL, optional model env vars (LMSTUDIO_MODEL*, RALPH_MODEL) and RALPH_MAX_ITER — these are proportional to the purpose. Important caveat: LMSTUDIO_URL can be set to any URL, which would redirect all code and diagnostics to that endpoint; that is expected behavior but a potential data-exfiltration vector if misconfigured.
Persistence & Privilege
always:false and no special persistence. The skill does not modify other skills or system-wide agent config and does not request elevated privileges. Autonomous invocation is allowed (platform default) but not excessive here.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install ralph-wiggum-loop - After installation, invoke the skill by name or use
/ralph-wiggum-loop - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
- Initial release of the Ralph Wiggum - AI Loop Technique skill.
- Enables automated, iterative code or text improvement using an LLM via a Generator > Critic > Fixer > Verifier loop.
- Provides a bash script (`ralph-loop.sh`) with options for file, inline code, custom prompts, iteration limit, output format, and model selection.
- Includes a modular architecture and Python API for direct integration.
- Comprehensive documentation with usage examples, troubleshooting, and detailed prompt engineering.
Metadata
Frequently Asked Questions
What is Ralph Wiggum Loop?
Iteracyjnie doskonali kod lub tekst AI, wykrywając i naprawiając błędy, optymalizacje, bezpieczeństwo i styl w maksymalnie trzech krokach. It is an AI Agent Skill for Claude Code / OpenClaw, with 82 downloads so far.
How do I install Ralph Wiggum Loop?
Run "/install ralph-wiggum-loop" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Ralph Wiggum Loop free?
Yes, Ralph Wiggum Loop is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Ralph Wiggum Loop support?
Ralph Wiggum Loop is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Ralph Wiggum Loop?
It is built and maintained by nerua1 (@nerua1); the current version is v1.0.0.
More Skills