Skip to content

Security: PavloSEO/seotools

SECURITY.md

Security policy

Reporting a vulnerability

Please do not open a public issue for a suspected vulnerability. Use GitHub private vulnerability reporting:

https://github.com/PavloSEO/seotools/security/advisories/new

Include the affected command or MCP tool, version, minimal reproduction, expected impact, and any safe mitigation you have already tested. Do not include real credentials or client crawl data.

Security model

SEOHEAD Tools is a local CLI and stdio MCP server. It has no hosted control plane, user account, telemetry collector, automatic updater, or inbound network port.

The main trust boundaries are:

  • remote HTML, XML, headers, redirects, robots.txt, and provider responses are untrusted data;
  • URL tools block private and non-public network targets unless explicitly enabled for an authorized staging or intranet environment;
  • URL fetches and direct TLS probes resolve once, reject non-global addresses by default, and connect to the vetted address while retaining the original hostname for SNI and certificate verification, so a resolver cannot answer the check and the connection differently;
  • addresses that carry a non-public destination inside a globally-scoped form — NAT64, 6to4, IPv4-mapped — are judged by the address they carry, not by the wrapper;
  • file-producing tools should run inside a dedicated working directory or mounted container volume;
  • operations that probe paths, rewrite files, verify bots over DNS, or spend provider credits require explicit input;
  • credentials are loaded from the local environment/configuration and should never appear in prompts, issue reports, transcripts, or committed files.
  • authenticated crawl headers use host-bound environment references; the global http.headers setting rejects credential-bearing header names.

Supported versions

Security fixes target the latest 3.x release and the current main branch. Versions older than 3.0.0 are not supported. Include the affected version and commit SHA in reports.

Out of scope

  • findings that require deliberately disabling documented guardrails;
  • third-party service outages or inaccurate provider datasets;
  • vulnerabilities already fixed by a current direct dependency release;
  • SEO disagreements without a confidentiality, integrity, or availability impact.

There aren't any published security advisories