Generate adversarial inputs and log discrepancies between target systems and oracles.
npx skills add https://github.com/gaasher/Agent-Loop-Skills --skill red-team-gaasher
Please help me install this Agent Skill. Skill: red-team Source: https://github.com/gaasher/Agent-Loop-Skills/tree/main/loops/red-team Command: npx skills add https://github.com/gaasher/Agent-Loop-Skills --skill red-team-gaasher
This Skill automates the process of adversarial testing to identify security vulnerabilities in guardrails, classifiers, prompts, or APIs.
Run the red-team skill with the target command and oracle command to start the adversarial testing loop.
High-intent search queries and answers about installing and using this skill.
Adversarial testing for AI security vulnerabilities is automated by generating adversarial inputs and running them through your target system to identify discrepancies against a ground-truth oracle. This skill requires Python 3.9+ and specific command-line tools to execute the continuous testing loop.
Testing the robustness of guardrails and classifiers is best achieved by providing a set of inputs to a target command and oracle command. The system continuously runs tests, generates adversarial inputs, and logs behavioral discrepancies over time to evaluate security.
You can use this to run vulnerability assessments on your own API by executing the red-team skill with your target command and oracle command. It generates adversarial inputs and observes how your API behaves over time against a ground-truth oracle.
You need Python 3.9+ to run automated adversarial tests because the skill depends on it for execution. Additionally, specific command-line tools and the harness.py script are required to generate inputs and compare target system discrepancies against the oracle.
The feedback loop in automated vulnerability assessments works by continuously running generated adversarial inputs through the target system. It then logs discrepancies between the target system's behavior and the ground-truth oracle to identify security vulnerabilities.
heman1033p-createFetch filtered schedule events from an Oracle API and output a Markdown table.
lebsralIdentify AI safety vulnerabilities through automated adversarial testing with DSPy.
ArabelaTsoGenerates randomized and edge-case Python test code for fuzzing software components.
barrydobsonAutomate secure CI/CD pipelines with GitHub Actions workflows.
