Hi, we at MDN are considering adopting ESLint for our code examples.

We have got some initial success at fixing thousands of syntax errors / stylistic issues. However, we have some code blocks that intentionally demonstrate bad syntax:

```js example-bad
// You can't use a try...catch without braces
try doSomething();
catch (e) console.error(e);
```

We decided to not use eslint-skip anywhere (or at least as sparsely as possible), so that external contributors don't get confused. On top of that, we don't want to duplicate metadata: since example-bad is already used by the platform to render a code block with a cross, we think the linter should accommodate that convention.

Therefore, I want to propose a feature that ignores code blocks by meta string—currently, the parser takes advantage of glob patterns to differentiate code blocks, so adding a Markdown-specific option could be tricky. I don't have a solid API design on top of my head, but maybe we can add the meta string as part of the language extension (e.g. *.example-bad.js)?