Keep WordPress shortcodes out of derived excerpts

[ssavutu]

A crossword post's whole body is a single [puzzleme ...] shortcode. deriveExcerpt strips HTML tags, which leaves brackets alone, so the derived excerpt was the shortcode source — basepath, embed ids, attribution markup and all — and the public site printed it under the headline on the Comics & Puzzles listing:

"What's on Tea-V?" — [puzzleme basepath='https://puzzleme.amuselabs.com/pmm/' set='4d204c…' id='dc2b9486' attribution='Made by Coco Li using the online <a href=…>cross word generator</a> from Amuse Labs' type='crossword']

Shortcodes are now dropped before tags are stripped, since attribute values carry HTML of their own (the puzzleme attribution is a link) that would otherwise be torn out from inside the brackets first. Two shapes match, both narrow enough to leave editorial brackets like [sic] and [Editor's note] alone:

• the shortcodes this corpus carries, named explicitly (puzzleme, caption, gallery, embed, playlist, audio, video);
• any bracketed token carrying attributes, which is what makes it a shortcode rather than prose.

Scope: this governs excerpts derived from here on. Rows that already store a shortcode excerpt are cleared directly, and DrexelTriangle/Scalene#77 strips on render so nothing else in the corpus can leak the same way.

Tests: four deriveExcerpt cases for the shortcode shapes plus one asserting editorial brackets survive.

🤖 Generated with Claude Code