Skip to content

Back to the catalog

microsoft/RAMPART
github.comrepo

microsoft/RAMPART

RAMPART (Risk Assessment & Measurement Platform for Agentic Red Teaming) is a Microsoft-built pytest-native framework for writing and running safety and security tests against agentic AI applications. It structures tests around adversarial attacks, benign failure modes, and harm categories, using evaluation-driven assertions integrated directly into pytest workflows.

TL;DR Microsoft's pytest-based framework for red-teaming and safety-testing AI agents.
Open original ↗
Document

What it is

RAMPART is a Microsoft framework for writing safety and security tests for agentic AI applications as pytest tests. The README adds one thing the catalog summary lacks: an onboarding skill called RAMPARTner, which runs inside a coding assistant.

What to do

The skill is a self-contained file at rampart/skills/rampartner/SKILL.md. It ships in both the repository and the Python package, so it is available after installing the package. According to the README, it helps the user:

  • agree on what a first test should prove,
  • build the integration with their application,
  • run that test,
  • choose follow-up tests and payloads grounded in their code,
  • decide whether to extend their own project or request a framework feature.

An engineer who wants to try RAMPART can load that file into their coding assistant and describe their agent's use case, instead of starting from the test API.

Caveats

The README excerpt gives no install command, version, supported agent frameworks, harm category list, or example test. It does not say which coding assistants can load the skill. Judging how well the framework fits a given stack needs the repository's docs and code.

Technologies named here

Related in AI Security