The Leak™ Framework is an evaluation methodology — structured methods, scoring rubrics, and prompt-pair templates for testing AI agent constraint integrity under operational pressure.

This repository contains methodology documentation only. It does not contain executable attack code, exploit tooling, active scanning infrastructure, or vulnerability disclosure material.

Leak evaluations are designed for use by authorized evaluators testing systems they own or have explicit written permission to evaluate. Use of this methodology against systems without authorization is outside its intended scope and may constitute unauthorized access under applicable computer fraud and abuse laws.

If you identify a flaw in the framework — a scoring calibration gap, a surface definition issue, or a logical inconsistency — please open a public GitHub Issue describing the problem and your reasoning. Methodology discussions benefit from open review.

If you identify a scenario prompt or example in this repository that could be misused to cause harm, do not open a public GitHub Issue. Use GitHub's Private Security Advisory feature instead:

1. Go to the Security tab of this repository
2. Click Report a vulnerability
3. Describe the concern privately

This ensures the issue is reviewed before it is publicly visible.

Tracy Pertner / Pertner Logic
linkedin.com/in/tracypertner
tpertner.github.io