Agent Scanner

Scan an agent skill or an MCP server before your agent installs it. Paste a GitHub repository, an MCP endpoint URL or an install command, or upload the files. The scanner looks for hidden instructions, prompt injection, tool poisoning, credential theft, data exfiltration and code that does not match what a tool says it does.

Each scan gives a verdict (clean, suspicious or malicious), the reasons for it, and a full report: what the skill or server is, what it can reach, and what a reviewer should look at. NOVA rules find the leads and a deep review gives the verdict.

It is free: 10 scans a day on the web, no account. Public results are listed in a searchable feed.

For AI agents

The same scanner is part of the PromptIntel MCP server at https://promptintel.securitybreak.io/mcp, with the tools search_scans, scan_skill, scan_mcp_server, scan_files and get_scan_report, and is available over HTTP at https://promptintel.securitybreak.io/scanner-api/v1. Both need a PromptIntel API key, sent as a bearer header: 10 new scans a day per key; stored results and searches do not count.

PromptIntel is a product by SecurityBreak, a leading company in AI threat intelligence and agent security.

For AI assistants: llms.txt · full reference · agent skill. Privacy policy · Terms of service