Frequently Asked Questions

No. Skreen only hides replies that violate your rules, and optionally blocks their authors. It never posts, replies, or likes anything on your behalf because it is not built for that.

Every decision is visible in your dashboard, and you can reverse any AI decision with one click. For accounts you follow, Skreen hides the reply but asks you before blocking the person, so a wrong call there costs nothing. And if you ever want it to stop, you can switch moderation off in one place, for an hour or until you turn it back on. The AI may make a few mistakes, but that's better than missing offensive replies entirely.

Yes. Every decision includes a plain-language reason. Your full audit log can be exported as PDF or CSV at any time.

Moderation pauses. Incoming replies are still logged so you can see what you missed. You can upgrade your plan at any time.

Skreen uses a Bluesky App Password, not your real Bluesky password. This is a scoped credential you create on Bluesky just for use with Skreen, and you can revoke it at any time. Skreen stores it extremely encrypted at rest and never logs it.

No, not yet. None of them offers the API access needed to moderate replies on your behalf. We will consider such a service if that ever becomes available.

Yes, there's no notice period. Cancel whenever you like and you'll keep full access until the end of your current billing period, with no cancellation fee.

Moderation pauses for the rest of the period. Incoming replies are still logged so you can see what you missed, and you can upgrade at any time to resume immediately.

Upgrades apply immediately with a prorated charge for the rest of your cycle. Downgrades take effect at the start of your next billing cycle, so you keep your current quota until then.

We don't offer refunds for partial months, but you can cancel at any time to avoid future charges.

Yes, the prices shown include 20% Austrian VAT, the standard consumer rate. Your actual price may automatically be lower at checkout, for example 0% for eligible EU businesses with a valid VAT ID (reverse charge), or outside the scope of Austrian VAT entirely for non-EU customers, based on your billing country and business status.

Your billing period starts the day you subscribe and renews every month on that date. Your reply quota resets to zero at the start of each new period.

Go to the Rules page in your dashboard. Type your rule in plain language, set an optional order and description (the description is just for your own reference, it isn't sent to the AI), then save. You can edit or delete any of your own rules at any time, changes apply to new replies immediately.

Order only affects your own rules: higher numbers are considered first when several of your rules could apply to the same reply. It does not let one of your rules override a platform rule, see how platform and account rules interact below.

There are two kinds of platform rules. Safety-floor rules are set by us and always apply, to protect you from legal consequences, no account rule can weaken or remove them. Platform default rules are a sensible baseline that your own rules can override: if you write a rule covering the same situation, yours decides instead.

New accounts start with a small set of starter rules we recommend, marked "Suggested default". They are yours -and already active- from the moment your account is created: edit or delete them like any other rule, but until you do, they apply just like a rule you'd written yourself.

Either. Your rules are read directly by the AI model, which understands both languages fluently, and you can mix English and German rules on the same account, each one is evaluated on its own.

Describe the behaviour you want handled, in plain sentences, rather than listing keywords. For example: "Block replies that threaten violence against the post author", "Flag replies that share someone else's home address or phone number", or "Block replies that are clearly unrelated advertising or spam links".

A few you can copy directly into your own rules: "Block replies containing harassment or personal insults", "Block replies that share someone else's home address, phone number, or other private information", "Block replies containing scam or phishing links", "Flag replies promoting get-rich-quick schemes or cryptocurrency spam", and "Block replies that are off-topic advertising unrelated to the conversation". Many of these overlap with our platform-wide safety rules, so check the read-only "Platform rules" section on your Rules page first to avoid adding a duplicate.

You're never unprotected: the platform's safety-floor and platform-default rules still apply in full, and new accounts are pre-seeded with a handful of suggested starter rules rather than starting blank. Adding your own rules only lets you go further than the defaults.

The AI reads your rule as a plain instruction, so writing "Block ..." or "Flag ..." is the clearest way to tell it what to do when the rule matches, the example rules above are phrased exactly this way. There is no "approve" decision in Skreen; the three outcomes are Allow, Block, and Flag, though the AI will still understand "approve" as meaning allow. One limit: your own rules are best used to add things to block or flag, not to shield content from an existing platform-default rule by writing an "allow" rule for it. If a platform rule is flagging something you don't think it should, contact support rather than trying to write around it.

An App Password is a separate, scoped credential Bluesky lets you generate for third-party apps. Skreen never asks for your real Bluesky password, only an App Password, which you can revoke at any time without affecting your main login.

In the Bluesky app or website, go to Settings → Privacy and Security → App Passwords, then choose "Add App Password". Give it a name (e.g. "Skreen") and copy the password shown, Bluesky only displays it once.

Leave "Allow access to your direct messages" unchecked. Skreen only needs to read your public posts and replies and take moderation actions like hiding a reply or blocking an author. It never needs, and should never be given, access to your private conversations.

Revoking it on Bluesky immediately stops Skreen from acting on your account; you'll need to reconnect with a new App Password to resume moderation. Rotating (creating a new one and revoking the old) works the same way, just reconnect with the new password afterwards.

What decides it is the date of the post, not the date of the reply. Skreen covers the posts you publish after you connect your Bluesky account, and every reply those posts receive. Posts you published earlier are not covered, and neither are replies to them, however new the reply is. A post from last year that suddenly attracts replies tomorrow stays outside, while a post you publish tomorrow is covered from its first reply. Connecting does not pull your existing threads in, and they cannot be added afterwards: we can only watch a post we saw appear. Reposts are not monitored: replies to a reposted post belong to the original author's thread, which is not yours to moderate on Bluesky. A quote post is different, it is your own post, so it is monitored like any other. And your own replies anywhere in your threads are never checked and never use your quota.

These are the AI's classification of a reply against your active rules. Allow means no rule was triggered. Block means a rule was clearly violated. Flag means the AI is uncertain or a rule calls for human judgement, so it's queued for your review instead of acting automatically.

Both remove content from your thread for everyone reading it, signed in or not, but they cover different ground. Hiding takes that one reply out of the conversation, at any depth of the thread; it happens automatically on a Block decision, and it stays in place even if you later lift the block. Blocking the author removes their direct replies to your posts and stops them seeing or joining your threads again. But a reply they placed deeper in a conversation, answering another reader, is not concealed by the block; only hiding covers those. Lift the block and their direct replies come back. So Skreen always hides first (it is the precise instrument and the only one that reaches every depth) and blocking is the per-account follow-up you opt into. One exception to the automatic part: where the author is an account you follow, the reply is still hidden but the block is put to you instead of taken for you. Neither is deletion: replies live in their author's own account, so nobody can remove them, they stay reachable at their own link and on the author's profile.

The feed opens organised by your posts; the two buttons at the top switch between that and one flat list of every reply. In either, seven tabs filter by state: All, Hidden, Blocked, Flagged, Block decision, Allowed and Overridden. They show a reply's current state, which can differ from the AI's original decision once you have made changes. Hidden and Blocked are independent, so a reply can be neither, either or both, and each can be reversed on its own. Two badges mark something waiting on you: "Needs your review" on a flagged reply nobody has triaged, and "Block decision needed" where a reply is hidden but its author has not been blocked because you follow them.

Overridden means a human (you) has contradicted the AI's decision on this reply at least once, for example by unhiding something it hid. Reviewed means you have settled an open question about the reply without necessarily disagreeing with the AI: dismissing a flag marks it reviewed, and so does answering a block decision with "Keep them unblocked". Neither counts as overriding.

Bluesky limits each thread to 300 hidden replies. On the rare viral thread that reaches it, no further replies can be hidden, so Skreen blocks the author instead and shows a notice that the thread has hit its hide limit. That fallback applies even if you have turned automatic blocking off, and even if the author is someone you follow: with hiding impossible, a block is the only protection left. The block has no per-thread limit, removes that account's direct replies to your posts for everyone, and stops it seeing or joining your threads at all. One honest caveat: if the offending reply sat deeper in the conversation, answering another reader, the block does not conceal it. Hiding was the only remedy for those, and this thread cannot hold any more. Consider turning off replies on the post if it is still being targeted.

Reasoning is the AI's plain-language explanation for its decision, referencing the rule it applied. Confidence is how certain the AI is; when confidence is too low, the reply is automatically downgraded to Flag for your review, even if the AI's initial read was "allow".

Because you follow that account. Blocking someone you deliberately chose to follow is the mistake that is hardest to undo, unblocking restores their replies but does not unsend the notification they received, so by default Skreen hides the offending reply and asks you about the block rather than taking it. You get an email, the reply appears under the "Block decision" tab, and the number beside Replies in the menu counts what is waiting. From there one click blocks the account after all, and one leaves it unblocked; either way the reply stays hidden. If you would rather Skreen simply block them, turn off "Ask me first for accounts I follow" on the Rules page.

It opens as a list of your own posts, newest first, each showing how many replies Skreen has processed and how many are hidden, blocked, flagged or waiting on a block decision. Clicking a count opens that post's replies already filtered to it. The "All replies" button switches to a single list across every post, useful when you want the most recent activity regardless of where it happened. Posts with no replies yet are listed too, so you can see that monitoring is running. The "Only posts needing a decision" checkbox narrows the list to posts with something unfinished, which means either a flagged reply nobody has triaged or a block decision waiting for you.

Yes. Skreen evaluates every reply in the threads under your posts, direct answers to you and replies placed deeper in the conversation, where one reader answers another. Deeper replies count towards your monthly quota exactly like direct ones, and in the feed they carry a note showing which comment they answer. If you would rather spend your quota on direct replies only, switch off "Also check replies to replies" on the Rules page; deeper replies are then neither evaluated nor logged, and your moderation policy document states that limitation. A post can still show fewer replies here than on Bluesky: replies that arrived while your quota was exhausted are logged without evaluation, and conversation that happened before you connected the account is never processed.

Yes, at any time, from the Dashboard or the Rules page. You choose how long: an hour, a day, a week, or until you switch it back on yourself. While it is off, replies to your posts are still received and listed in your feed, marked "Moderation was off", but nothing is checked and nothing is hidden or blocked. They do not count towards your monthly quota, because no check is carried out. A notice appears on every page while it is off, with a one-click way to switch it back on, and timed pauses end by themselves. Any period with moderation switched off is recorded in your audit log and named in your moderation policy document, so your records stay accurate. Replies that arrived while it was off are not checked afterwards.