ToolExperimental

mantis

Name: mantis
Rating: 3.3 (2519 reviews)
Author: google

by google

Security-review skills for AI coding agents to find, reproduce, and patch vulnerabilities

Python

Updated Oct 24, 2026

2.5k

Stars

216

Forks

Commits/Week

Commits/Month

View on GitHub

Overview

Provides a modular toolkit of security-review skills that let AI coding agents autonomously find, reproduce, and patch vulnerabilities. security-review skills Composes language-model-driven analysis, exploit reproduction steps, and automated patch suggestions into reusable agent skills you can wire into multi-agent workflows. Includes test harnesses and CLI tooling to run security passes against repositories and generated code.

Avoid common pitfalls

Learn what failures to watch for

Key Benefits

As agent chains take on more autonomous development tasks, knowing whether an agent introduces or misses vulnerabilities becomes critical to trust. Mantis gives agent ecosystems concrete security-oriented skills and observable checks, so reputation can include vulnerability detection and remediation performance. Treating security review as a first-class agent capability helps surface durable trust signals and reduces blind spots in agent-to-agent handoffs. To standardize cross-agent exchanges, consider the Model Context Protocol Model Context Protocol (MCP). trust signals

When to Use

Researchers and engineering teams building or evaluating coding agents who need automated security review skills and reproducible vulnerability checks. For those teams, automated security review skills enable reproducible workflows across agents.

Use Cases

Automating security reviews of code changes using LLM-driven exploit reproduction
Adding reproducible vulnerability detection and patch suggestion skills to agent pipelines
Benchmarking agents on security-specific tasks to build an agent track record