mirror of
https://github.com/zoo-apps/fund.git
synced 2026-10-30 14:57:02 +00:00
zeekayandHanzo Dev
87e8ef3b92
An audit turned this on its author. bin/cites.mjs reported "36 sources, 0 unreachable" while thirteen fabricated citations sat in the file it was scanning, and both reasons were mine. nature.com was on the skip list. This file's own opening paragraph names nature.com/articles/ai-conservation-2024 as the fabrication it was written to catch, and then line 38 told it not to look. I put it there after a real-looking DOI returned 404 and I read that as bot-blocking. It is not: Nature answers 200 for a real article and 404 for an invented slug. Five fabricated Nature citations were standing behind that exemption. The URL pattern stopped at `?`. Every youtube.com/watch?v=<id> was therefore tested as a bare /watch, which answers 200 and names no video. Eight invented ids — ocean-edna, ml-wildlife, zenlm-africa — passed. A green check standing in front of the defect it was written for is worse than no check. So: the exemption list now carries a control URL per host and refuses to run at all if one answers 200, because an exemption that is not earned hides precisely what this looks for. worldwildlife.org and coindesk.com keep theirs — both 403 and 429 their own front pages. YouTube is asked through oEmbed rather than by status or by body. The status is 200 for anything; the body says "unavailable" on pages that play fine, so matching it failed real videos — the crying-wolf failure this file's own comment warns about, which I then wrote. oEmbed returns 200 for a video that exists and 400 for one that does not. With the check honest, the thirteen it had been hiding are removed. Co-authored-by: Hanzo Dev <[email protected]>