[jooyoungseo]

Description

Closes #15023.

The dashboard filter lifts a cell's raw HTML output out of its card when it takes the output for bslib components. It tested for them with text:match('bslib-'). In a Lua pattern - is a quantifier (zero or more of the preceding character, as few as possible), so the pattern matched any text containing bsli, for example:

• the word "subslide";
• bslib with no hyphen.

Such output left its card, and a card with nothing else in it was dropped, title included.

The pattern is now 'bslib%-'. It matches only the literal bslib-, the prefix bslib uses in its classes, ids, attributes and custom elements such as <bslib-tooltip>.

One real component changes. py-shiny's ui.accordion() without an id names its elements bslib_accordion_* and contains no bslib-, so it was lifted only through the loose match. It now stays in its cell's card, with the card's title, like any other output. Two related outputs are still lifted:

• R's bslib::accordion(), whose ids are bslib-accordion-*;
• a py-shiny accordion given an id, which adds the class bslib-accordion-input.

Not changed here:

• Dashboard: bslib::value_box() ignores #| expandable: false and shows Expand button #14719, where cell options are not applied to real bslib output that is lifted;
• c:match("^column-") in quarto-pre/parsefiguredivs.lua, which has the same unescaped - but is anchored.

Tests

New tests/docs/smoke-all/2026/10/11/issue-15023.qmd, a dashboard with three Python cells:

• the outputs "A subslide of the deck" and "Cards are built by bslib" must stay in the card built from their cell;
• an output with a bslib- class must still be lifted out of its cell's card.

The selectors are scoped to .card.cell, the card built from a cell. Without the fix all three outputs are lifted, so no cell card is left. The dashboard then wraps the page's loose content, those outputs included, in a card of its own without .cell. A selector on .card alone, such as .card .test-subslide, would match inside that card and pass.

Locally:

• Without the fix, the new test fails on .card.cell .test-subslide.
• With the lift disabled (its condition forced to false), it fails on .card.cell .test-lifted, so the test also catches real bslib output left in its card.
• With the fix, it passes, and so do all 49 smoke-all tests under docs/smoke-all/dashboard/.
• A dashboard whose cell outputs bslib::value_box() renders byte-identical HTML with and without the fix.

Checklist

I have (if applicable):

• referenced the GitHub issue this PR closes
• updated the appropriate changelog in the PR
• ensured the present test suite passes: the smoke-all tests under docs/smoke-all/dashboard/, run locally
• added new tests
• created a separate documentation PR in Quarto's website repo and linked it to this PR: not applicable, as no option or documented behavior changes

AI-assisted PR

• AI tool used: Claude Code
• Codebase grounding: local clone of quarto-cli, including its .claude/rules and llm-docs
• Human review: I have reviewed, tested, and verified the AI-generated content before submitting.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[posit-snyk-bot]

✅ Snyk checks have passed. No issues have been found so far.

Status	Scan Engine	Critical	High	Medium	Low	Total (0)
✅	Open Source Security	0	0	0	0	0 issues
✅	Licenses	0	0	0	0	0 issues

💻 Catch issues earlier using the plugins for VS Code, JetBrains IDEs, Visual Studio, and Eclipse.

[cderv]

Thank you !