Adversarial Python Packages

alibi-detect

Algorithms for outlier, adversarial and drift detection

pyabsa

Sentiment Analysis, Text Classification, Text Augmentation, Text Adversarial defense, etc.;

pdfhell

Adversarial PDFs that break AI document readers. Procedural ground truth, not LLM-as-judge.

redteam-arena

AI vs AI adversarial security testing platform. Red team agents attack, blue team agents defend. Fully automated.

cospgd

The official repository for CosPGD: a unified white-box adversarial attack for pixel-wise prediction tasks.

preclinical

Preclinical is an open-source healthcare AI safety testing platform. It simulates realistic adversarial patient interactions against your agent, stores transcripts, and grades outcomes against safety rubrics.

adversarial-solver

Two AIs argue until your content is right. Generator writes, Critic attacks, loops until PASS. Built for teams tired of fixing AI slop.

palisade-scanner

Scan web content for prompt injection, hidden instructions, and adversarial content targeting AI agents

kuang

Human-in-the-loop adversarial code-review panel — a bounded loop of reviewer agents that try to break your change, with a ruin-class circuit-breaker.

ocarinalabs-harbor-ext

Adversarial probes for Harbor benchmark integrity. Ocarina Labs' Harbor extension — null-agent, output-echo, judge-injection, and verifier-tamper probes that flag broken benchmarks before they ship.