Skip to content

Security: OutSystems/outsystems-mcp

Security

SECURITY.md

Security Policy

Scope

This repository ships the OutSystems MCP integration manifests and skill content for AI assistant harnesses (Claude Code plugin, Kiro Power, and a generic SKILL.md for other harnesses). The actual MCP server is hosted by OutSystems and is not part of this repo.

Reports are routed differently depending on what you found:

What you found Where to report
A vulnerability in the manifests, skill content, install snippets, or any other file in this repository GitHub private security advisory on OutSystems/outsystems-mcp (see below)
A vulnerability in the hosted MCP server, the OAuth flow served by OutSystems, the OutSystems platform itself, or any service reachable via the MCP transport OutSystems Trust Center: https://www.outsystems.com/trust/

If you're unsure which bucket your finding falls into, default to the GitHub private advisory on this repo and we'll route it.

Reporting a vulnerability in this repository

Please do not open a public GitHub issue. Public issues are visible to everyone and can put users at risk before a fix ships.

Use GitHub's private security advisory flow:

  1. Go to https://github.com/OutSystems/outsystems-mcp/security/advisories/new.
  2. Describe the issue, including:
    • The affected file(s) and commit SHA or release version.
    • Reproduction steps.
    • Impact — what an attacker could do, and under what preconditions.
    • Any suggested fix or mitigation, if you have one.
  3. Submit. Repository maintainers receive a private notification.

What to expect

  • Acknowledgment: we'll confirm receipt as soon as we can, typically within a few business days.
  • Status updates: we'll keep you posted as the investigation progresses.
  • Fix priority: driven by severity and impact, on a best-effort basis — critical findings get the fastest turnaround we can manage.
  • Disclosure: once a fix is released, we publish a GitHub security advisory (with a CVE if applicable). Reporters are credited unless they request anonymity.

Out of scope

  • Issues that require a compromised local machine, a malicious AI assistant harness, or the user pasting an attacker-controlled install snippet. Those are the harness's threat model, not this integration's.
  • Findings against third-party dependencies that this repo does not customize. Report those upstream.
  • Best-practice opinions, missing rate limits, or non-vulnerability concerns. Open a normal issue or PR instead.

Supported versions

Only the latest release tagged on main receives security fixes. We do not back-port to older plugin or Power versions; bump to the latest version to pick up fixes.

There aren't any published security advisories