New
Introducing React Bench, see how different models perform on React code
Menu
What Is React Doctor?
Quickstart
Install for Coding Agents
How to Fix Issues
GitHub Actions Setup
Updating CI
Other CI Providers
Migrating to CI v2
Config Files
ESLint & oxlint Plugins
Add Custom Rules
CLI Reference
GitHub Action Reference
Node.js API
Changelog
Modified MIT License
Terms of Service
Privacy Policy
Data Use
Security
Support
Rules reference

react-doctor/mdx-ssr-execution-risk

Compiling untrusted MDX with the full pipeline runs attacker-supplied JSX and expressions on your server, which can lead to code execution.

  • Category: Security
  • Severity: warn
  • Source: oxlint-plugin-react-doctor
  • Framework: global
  • Enabled when: production source files (.js/.ts/.jsx/.tsx, etc.); test/build/doc/generated paths skipped
  • Tags: security-scan
  • Default: Enabled

Validation prompt

Use this to decide whether a fired diagnostic is real or a false positive.

Fires when an MDX library surface (@mdx-js/mdx, next-mdx-remote, or MDXRemote/compileMDX/evaluateMdx) is followed within ~700 chars by an untrusted-source or dangerous-flag marker: repo/customer/tenant/user-content|markdown|mdx|input|provided|generated|submitted/untrusted/searchParams/req./request./fetch(/prisma./db./database/rehypeRaw/allowDangerousHtml. Comments are stripped first. FALSE POSITIVE: the file compiles its own first-party, build-time MDX (docs/blog content from the repo's own content/ directory) with no attacker-controlled source — generic library usage co-occurring with a word like content is not itself a vulnerability; a bare evaluate/compile not tied to an MDX surface is also excluded.

Fix prompt

Use this once validation confirms the diagnostic is real.

Never compile or evaluate attacker-controlled MDX with the full pipeline. For untrusted content, use a constrained compiler: disable JSX, expressions, and raw HTML (do not enable rehype-raw or allowDangerousHtml), render through a sandbox with an allowlisted component set, and prefer plain Markdown that cannot run code. Do not cache one tenant's rendered output where another tenant can read it.

More Security rules from the rules reference:

  • react-doctor/nextjs-no-side-effect-in-get-handler: Move the side effect to a POST handler and use a <form> or fetch with method POST — GET requests can be triggered by prefetching and are vulnerable to CSRF
  • react-doctor/no-danger: Flag dangerouslySetInnerHTML — raw HTML injection bypasses JSX escaping and invites cross-site scripting.
  • react-doctor/no-eval: Use `JSON.parse` for serialized data, `Function(...)` (still careful) for trusted templates, or refactor to avoid dynamic code execution
  • react-doctor/no-secrets-in-client-code: Move secrets to server-only code. Public client environment variables are bundled into browser code and must not contain secrets
  • react-doctor/nosql-injection-risk: Building a NoSQL query from raw client input lets an attacker inject operator-shaped keys or `$where` code and read or alter data they should not.