Wikipedia:Edit filter noticeboard#Special:AbuseFilter.2F29

While patrolling the edit filter log, I noticed that ClueBot III was unable to create an archive because it tripped filter 1,347, attempting to archive content that contained protection template markup; see Special:AbuseLog/40769169. This wasn't the first time it has failed; ClueBot III's [https://en.wikipedia.org/wiki/Special:AbuseLog?wpSearchUser=ClueBot+III&wpSearchPeriodStart=&wpSearchPeriodEnd=&wpSearchTitle=&wpSearchImpact=0&wpSearchAction=any&wpSearchActionTaken=&wpSearchFilter=1347 filter log] shows several previous failed attempts. — MRD2014 (talk) 00:00, 12 May 2025 (UTC)

:I think we could just add !(bot in user_groups) to the top of the filter. – PharyngealImplosive7 (talk) 01:33, 12 May 2025 (UTC)

::{{pp should not be hit by the filter, bot or not. Adding !(added_lines irlike "{{pp") or something to the same effect is a good idea. Nobody (talk) 05:16, 12 May 2025 (UTC)

:::Done. I used a negative look-behind assertion. If there's additional text between the nowiki tag and the protection template sometimes then we'll need to use an additional condition instead, but I don't see any instances of that in the logs. I also exempted {{u|ClueBot III}} and {{u|Lowercase sigmabot III}}. Daniel Quinlan (talk) 06:40, 13 May 2025 (UTC)

Filter {{ef|231}} uses !(added_links to remove urls from the filter, but this has the effect that this edit doesn't get picked up by it, due to {{tl|coord}} (produces a geohack.toolforge url) from {{tl|Infobox French commune}} grabbing type as a optional parameter, which includes the population number. This has the effect that changes to | population can never be picked up by filter 231. I don't think this is the desired behaviour. Nobody (talk) 08:14, 12 May 2025 (UTC)

:From what I see this effects at least all 573k uses of {{tl|Infobox settlement}}. Nobody (talk) 08:23, 12 May 2025 (UTC)

::Maybe something like !(added_lines irlike "\|\s*population\s*=\s*\d") would work, but we may need more narrow regex. – PharyngealImplosive7 (talk) 16:00, 14 May 2025 (UTC)

:::I had to fix the regex above by removing the multiple/duplicate backslashes. Codename Noreste (talk · contribs) 05:06, 15 May 2025 (UTC)

I've noticed that filter {{abusefilter|1325}} ("Possible AI-generated text") has picked up edits where various idioms the filter is looking for happen to be used in the titles of references (for example [https://en.wikipedia.org/w/index.php?title=The_Black_Market_(Cork)&diff=prev&oldid=1289962144 here] and [https://en.wikipedia.org/w/index.php?diff=1290132873&oldid=1284591016&title=Serafim_Sepp%C3%A4l%C3%A4 here]). Please exclude these cases. Duckmather (talk) 02:27, 13 May 2025 (UTC)

:Done. I think this will increase the false negative rate somewhat because "must-see" and "must-visit" citations show up in some AI-generated articles, but there are just too many false positives due to citations right now. Daniel Quinlan (talk) 06:15, 13 May 2025 (UTC)

Has been set to disallow by me. I'm not opposed to making this a throttle and/or merge this into another filter (LTA ranges?).

Do not discuss specifics here, please use the mailing list. beef [talk] 08:08, 19 May 2025 (UTC)

I've noticed plenty of articles about schools in Wales use their Welsh-language names as titles, often containing the word 'ysgol' - :Category:Welsh-language schools has quite a few examples. Kids being what they are means these pages get their fair share of vandalism, so would it be worth including 'ysgol' in the titles targeted by filter 39 (for school-related article vandalism etc.) to make disruption to these pages easier to catch? Entranced98 (talk) 18:08, 19 May 2025 (UTC)

:{{done}} This seems like a helpful edition, and I don't see too many problems given how narrow the filter string is. EggRoll97 ^(talk) 03:10, 20 May 2025 (UTC)

::Thanks, much appreciated! Entranced98 (talk) 10:23, 20 May 2025 (UTC)