Pushed a follow-up commit (6704f82) to get this over the line on the review feedback:

1. Added a test (tests/cypress/integration/plugins/intersect.spec.js)

Added a .parent test that proves the modifier actually re-roots the IntersectionObserver at the parent rather than the viewport. The scroll container is pushed far below the fold (height: 200vh spacer), so a window-rooted observer would never fire — the count only increments when the content is scrolled into view within the container:

test('.parent observes the element relative to its parent, not the viewport', ...
    get('span').should(haveText('0'))
    get('#container').scrollTo(0, 250, {duration: 100}) // scroll within the parent, not the page
    get('span').should(haveText('1'))

I verified end-to-end in a real browser that the same markup gives 1 with .parent and 0 without it, so the test genuinely depends on the feature rather than passing incidentally. Ran the full spec locally — all 9 tests pass:

✓ .threshold (290ms)
✓ .parent observes the element relative to its parent, not the viewport (285ms)
9 passing (5s)

2. Fixed the docs (addressing both Copilot notes Josh flagged)

• Fixed the grammar in the example comment (element is moves inside it's parents → the element scrolls into view within its parent).
• Reworded the description, which previously said the default root "is the browser." IntersectionObserver defaults to the top-level viewport (root: null), so the copy now says the element is observed "against the browser's viewport" and explains that .parent sets the observer's root to the parent element.

I left the <a name="parent"> anchor in place since every other modifier section in this doc uses the same pattern.

(Pushed directly to the branch via maintainer edits — thanks @kevinruscoe!)

thanks!