OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Agent Security & Sandboxing

Agent Scan

Agent Scan is a Python command-line security scanner that discovers AI agent components and scans agents, MCP servers and skills for prompt injections and other vulnerabilities. It can inspect tools, prompts and resources associated with MCP servers, and supports both automatic discovery of supported agent configurations and scans of specified files or directories.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Agent Scan — OpenViST project profile in Agent Security & Sandboxing.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Agent Scan when reviewing an agent setup, MCP configuration or collection of skills for security issues. For untrusted MCP configurations, follow the project's sandboxing guidance and inspect consent prompts before allowing server contact.

Who it is for

  • Developers reviewing local AI agent installations and configurations.
  • Teams checking MCP servers and skills for security risks before use.
  • Security practitioners who want a CLI-based component scan in a local review process.

Documented uses

  • Discover supported agent configurations and installed components on a machine.
  • Scan an MCP configuration, a skill file or a skills directory.
  • Check agent components for risks including prompt injection, malicious content, suspicious downloads, credential handling and exposed secrets.
Before you choose

Requirements & limits

The documented scan setup requires a Snyk account and API token set in the SNYK_TOKEN environment variable. Installation options include uvx, which requires uv, and standalone binaries from GitHub Releases. The repository is licensed under Apache-2.0.

  • MCP scanning may execute configured commands or make outbound network requests; untrusted configurations should be scanned in a sandbox.
  • Interactive foreground scans request consent before contacting discovered MCP servers; background and push-key scans handle remote servers differently.
  • Raw CLI output is experimental and may change, including fields and risk indicators. The repository advises against production dependencies on specific output values.
  • Version 0.5.x is planned for deprecation; version 0.6 and later use risk-based output.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Scan an MCP configuration with uvx

Run Agent Scan against a specific MCP configuration file.

Before you start

  • A Snyk account and API token.
  • Set the token in the SNYK_TOKEN environment variable.
  • uv installed on the system.

Steps

  1. Choose the documented command for the installed CLI line: uvx snyk-agent-scan@latest ~/.vscode/mcp.json for version 0.6 and later, or uvx snyk-agent-scan@0.5.17 ~/.vscode/mcp.json for the documented v0.5.x example.

  2. Review the consent prompt carefully before allowing contact with discovered MCP servers.

  3. Use a sandbox when evaluating untrusted or third-party MCP configurations.

Expected outcome

Agent Scan inspects the specified MCP configuration and reports detected risks using the output format of the selected CLI line.

Guide limitations

  • Scanning configurations can execute configured commands or make outbound network requests.
  • CLI output fields and risk indicators are experimental and may change; do not build production workflows around specific output values.
  • Version 0.5.x is planned for deprecation.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
9 October 2026
Original website title
snyk/agent-scan · Official homepage title · retained in its source language
How we review sourcesSuggest a correction