RAMPARTner skill
The RAMPARTner skill guides a developer from idea to working pytest safety tests for an agentic app, explaining each step.
- 425 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The skill leads to running attack probes against an application; run only on systems you own.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/rampartnerDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The skill leads to running attack probes against an application; run only on systems you own.
This is third-party code. Review the repository files before installing.
What it does
RAMPARTner is a teaching skill from Microsoft's RAMPART package. The agent asks focused questions about the application, builds an adapter, writes and runs pytest safety tests, explains results and recommends further attacks and probes grounded in the project code. Before installing dependencies the skill asks whether to use the latest release or `main` and pins the choice. Every reply briefly explains the RAMPART concept behind the action.
Who it is for. Developers and testers of AI agents who need tests for attacks and harmful behavior and are new to the RAMPART framework.
Good fit when
- An agentic app is connected to RAMPART for the first time.
- The team needs to decide what a first safety test should prove.
- A test run's results are unclear and get explained by the agent.
- A team picks attacks and probes for its own scenario.
Not a fit when
- The code under test has no agentic part.
- A ready report is wanted with no developer involvement or questions.
- A ready attack suite without questions is wanted; the skill is a dialogue and sets its own pace.
Example request
Help me connect my app to RAMPART and write a first test for resistance to injected instructions.Limitations
The skill checks nothing itself; it steers an agent's work with the RAMPART package and depends on its version. The release or `main` choice stays with the user, and `main` may change APIs. Attack probes run against your application, so test only your own systems and staging. Recommendations for new tests rest on the project code the agent could see, so private modules it cannot read are not covered. The skill assumes Python and pytest.
How to disable. Delete the `rampartner` skill folder from the agent skills directory. Remove the RAMPART package and the tests it wrote from the project separately.
FAQ
Where is the skill?
In the repository at `rampart/skills/rampartner/SKILL.md` and inside the RAMPART Python package.
What is RAMPART?
A pytest-based framework for AI agent safety tests, covering attacks, benign failures and harm categories with evaluation-driven assertions.
Is a license fee involved?
No. The repository is MIT-licensed and the README lists no paid parts.
What does the skill ask first?
Which version to use: the latest release or `main`. The choice is pinned in the project's dependencies.
Does the skill explain results?
Yes. In every reply the agent briefly explains a RAMPART concept, and after a run it explains what the tests showed and which probes to add.
Discussion
No questions or reviews yet.
Related
SkillsA skills router for authorized reverse engineering, pentest, CTF and threat analysis: it picks a methodology for the task and checks available tools
A security scanner for MCP servers and agent skills that looks for prompt injection, secret leaks and swaps after approval.
A security research framework on Claude Code: static analysis, LLM validation of findings, exploits and patches
An AI agent behavior verifier plugin: it compares declared authority with code and settings and reports the gaps.
A security CLI for coding agents: secret scanning, command policies, pre-commit hooks, third-party skill auditing
An autonomous AI pentester that proves every finding: verified bugs ship a proof capsule you can replay
README in short
A pytest-native safety and security framework for agentic applications. The README sends newcomers to the RAMPARTner skill.