AB Test Framework

Compare models with A/B testing for selection

Security Scan

The skill's stated purpose (A/B testing) matches the general structure, but the actual implementation, examples, and metadata are inconsistent or incomplete — it appears unfinished and should be reviewed before use.

Purpose & Capability

Name/description match an A/B testing framework. However the code is a placeholder: it performs only input validation and returns sanitized params rather than running any A/B comparison logic. SKILL.md declares dependencies (openclaw/llm, stats-library) but code only depends on a stats-library in package.json and uses OpenClaw runtime helpers; openclaw/llm is not used. Declared complexity 'high' and 'priority: 5 (Critical)' are disproportionate to the actual footprint.

Instruction Scope

SKILL.md inputs and security guidance are reasonable, but examples contain inconsistencies (example passes test_prompts as '123' — wrong type). The runtime instructions don't request external files, secrets, or unexpected network endpoints. The skill logs full params (logger.info with params), which could capture sensitive identifiers — the SKILL.md does not explicitly warn about logging sensitive inputs.

Install Mechanism

No install spec is present (instruction-only install), and package.json only lists 'stats-library' as an npm dependency. No external download URLs or archive extraction are used. This is a low-risk install surface, though the declared dependency should be verified (stats-library from npm).

Credentials

The skill requests no environment variables, credentials, or config paths. It uses OpenClaw runtime helpers (logger, validator, notify) which are expected for a platform skill. The code conditionally calls sendAlert if params.alert_on_failure is set — no secrets are required for that.

Persistence & Privilege

always is false and the skill does not attempt to modify other skills or system settings. It does not request persistent system-level privileges.

What to consider before installing

This skill appears unfinished: index.js contains a placeholder implementation (no A/B comparison logic), SKILL.md and package.json list dependencies that are not consistently used, and example/test code contains type/parameter mismatches (e.g., missing or incorrectly-typed test_prompts). Before installing or using in production: 1) Review and complete the core A/B testing implementation and confirm how 'stats-library' will be used. 2) Fix tests and examples so they match required inputs (test_prompts is required). 3) Audit logging: the code logs full params — remove or redact sensitive model identifiers or prompt content if needed. 4) Verify the provenance of any npm dependency (stats-library) and ensure OpenClaw runtime helpers (openclaw/*) are trusted. If you cannot review/modify the code, avoid using this skill for sensitive workloads.

A/B Testing Framework

Description

Compare models with A/B testing for selection

Source Reference

This skill is derived from 20. Testing & Quality Assurance of the OpenClaw Agent Mastery Index v4.1.

Sub-heading: A/B Testing Frameworks for Model Selection

Complexity: high

Input Parameters

Name	Type	Required	Description
model_a	string	Yes	First model
model_b	string	Yes	Second model
test_prompts	array	Yes	Test prompts

Output Format

{
  "status": <string>,
  "details": <object>,
  "winner": <string>,
  "confidence": <number>
}

Usage Examples

Example 1: Basic Usage

const result = await openclaw.skill.run('ab-test-framework', {
  model_a: "value",
  model_b: "value",
  test_prompts: 123
});

Example 2: With Optional Parameters

const result = await openclaw.skill.run('ab-test-framework', {
  model_a: "value",
  model_b: "value",
  test_prompts: []
});

Security Considerations

A/B test security per Category 8; prevent test manipulation

Additional Security Measures

1. Input Validation: All inputs are validated before processing
2. Least Privilege: Operations run with minimal required permissions
3. Audit Logging: All actions are logged for security review
4. Error Handling: Errors are sanitized before returning to caller

Troubleshooting

Common Issues

Issue	Cause	Solution
Permission denied	Insufficient privileges	Check file/directory permissions
Invalid input	Malformed parameters	Validate input format
Dependency missing	Required module not installed	Run npm install

Debug Mode

Enable debug logging:

openclaw.logger.setLevel('debug');
const result = await openclaw.skill.run('ab-test-framework', { ... });

Related Skills

• model-routing-manager
• performance-benchmarker

• @param {string} params.model_a - First model
• @param {string} params.model_b - Second model
• @param {Array} params.test_prompts - Test prompts