PromptStrike is an adversarial red teaming framework for large language models. It automates the PAIR and TAP jailbreak algorithms from academic literature to measure how well an LLM resists manipulation.

Authorized use only. Only test LLMs you own or have explicit written permission to test. Unauthorized testing may violate the provider's terms of service and applicable law.

PromptStrike is intended for:

• Security research — measuring LLM robustness against known attack algorithms
• Red teaming engagements — with explicit authorization from the model operator
• AI safety evaluation — internal testing of your own deployed models
• CI/CD safety gates — automated pre-deployment regression checks

It is not intended for:

• Attacking production systems without authorization
• Bypassing safety measures on public LLM APIs in violation of their ToS
• Generating harmful content for any purpose other than measuring model behavior

If you discover a security vulnerability in the PromptStrike codebase (e.g. a dependency with a known CVE, an injection flaw in the dashboard, unintended credential exposure), please report it responsibly:

1. Do not open a public issue. GitHub issues are indexed and may expose the vulnerability before a fix is available.
2. Open a GitHub Security Advisory — this is private between you and the maintainer.
3. Include: a description of the issue, reproduction steps, and your assessment of impact.

I aim to acknowledge reports within 48 hours and publish a fix within 7 days for critical issues.

Dependencies are pinned with minimum versions in pyproject.toml. Run pip audit or safety check periodically to check for CVEs in the dependency tree.

pip install pip-audit
pip-audit