
Snyk Agent Scan
Snyk Agent Scan is a free open-source CLI used to discover and scan AI agents, MCP servers, and agent skills for security and engineering teams.
0 upvotes · 0 downvotes
No ratings yet
- Pricing
- Free
- Platforms
- LinuxWindowsmacOS
- License
- Open source
What it is
A Snyk command-line scanner, previously released as mcp-scan by Invariant Labs. It finds installed agent harnesses, MCP servers, and skills, then checks them for prompt injection, untrusted content, secrets, unsafe credential handling, and malicious skill code. Security analysis uses Snyk's Agent Scan API. Inspect mode skips that analysis.
What it helps with
- Inventorying local agents and their MCP servers or skills
- Flagging prompt injection and unsafe tool capabilities in MCP servers
- Finding secrets, unsafe credential handling, and malicious skill patterns
Who it's for
Security engineers reviewing agents and MCP setups on systems they own or are authorized to assess.
Worth knowing
Connecting to an MCP server can execute a local command or call a remote URL, so untrusted configs should be scanned in a sandbox. Raw CLI output is experimental and may change between releases.
Does analysis require a Snyk account?
Yes. The scan command needs a Snyk API token, or a push key, to use the analysis API.
How does it compare with garak, promptfoo, and Hermes Security?
In our view garak and promptfoo probe model and prompt behavior, and Hermes Security reviews code inside Hermes Agent, while Snyk Agent Scan checks agents, MCP servers, and skills already on a machine.
Discussion & reviews
0 comments
No contributions yet. Be the first to review or comment.