Responsible Disclosure Policy
Last updated: March 2026
Scope
This policy covers security vulnerabilities discovered by MCP Sentinel's automated scanning infrastructure across the MCP server ecosystem. We scan 21,000+ servers against 177 detection rules covering prompt injection, command injection, data exfiltration, supply chain attacks, and more.
Our Commitment
- We will not publicly disclose specific vulnerabilities with exploitable detail until the server author has been notified and given reasonable time to remediate.
- We publish aggregate statistics (e.g., “23% of servers have prompt injection risks”) without identifying individual servers.
- We display findings on server detail pages with remediation guidance — this is intended to help, not shame.
- Server authors can dispute findings they believe are false positives by contacting us.
Disclosure Timeline
What We Scan
MCP Sentinel's scanner performs passive analysis only:
- Tool descriptions and parameter schemas (no tool invocation)
- Source code from public GitHub repositories
- Package dependencies from npm/PyPI manifests
- MCP protocol metadata via
initialize+tools/list(read-only)
We never invoke tools on scanned servers. We never send test payloads, make authenticated requests, or interact with server-side resources.
Dispute a Finding
If you believe a finding on your server is a false positive, please open an issue on our GitHub repository with the server slug and finding details. We will review and update the finding status within 7 days.
Report a Vulnerability in MCP Sentinel
If you discover a security vulnerability in MCP Sentinel itself (our scanner, website, or API), please report it via GitHub Security Advisories on our repository. Do not open a public issue for security vulnerabilities.