Twitter

Your assistant can run X's own search syntax, so it can look for your company name, your product, your handle and the misspellings all at once, then read the posts rather than counting them. It can also pull the volume of matching posts by day, which is how you tell a normal week from something starting to move. Quote posts are a separate read and they are usually where the opinion is.

What you end up with. A written read of what was said about you this week, with the posts worth answering pulled out and linked.

Start with it bringing you the information and you deciding.

Never without. Know that recent search only reaches back seven days. Anything older needs the separate full archive action, and an assistant told to check the last month will quietly return three weeks of nothing.

Built on: Search recent tweets / Fetch recent tweet counts / Retrieve posts that quote a post / Look up post by id

The assistant can pull your recent posts by searching on your own handle, then read who liked each one and who reposted it. That gives you engagement by named account rather than as a number, which is more useful than it sounds: ten likes from people in your industry is a different result from ten likes from bots.

What you end up with. A short read of your recent posts ranked by real engagement, with the accounts that engaged named.

Start with it bringing you the information and you deciding.

Never without. Do not expect impressions. The post analytics action requires an X Enterprise plan and returns an error on Free, Basic and Pro, so engagement here means likes, reposts and quotes and nothing about reach.

Built on: Search recent tweets / Get tweets by IDs / List post likers / Get post retweeters

Your follower list is readable with full profiles, and so is anyone you name by handle, up to a hundred at a time. The assistant can cross-reference against who you already follow and read their recent posts, so instead of a list of avatars you get a description of who arrived and what they are actually interested in.

What you end up with. A described list of new followers worth knowing about, with what each of them posts about.

Start with it bringing you the information and you deciding.

Never without. Treat a low follower count and a new account as unproven rather than as fake. This is a read about real people, and having your assistant sort strangers into worth-knowing and not is a judgment you should be making yourself.

Built on: Get followers by user id / Get following by user ID / Look up users by username / Search recent tweets

Recent direct message events are readable, and the assistant can open a specific conversation to get the full thread rather than just the latest line. That is enough to work out which conversations end on someone else's message. It reads only; nothing gets replied to or marked as anything.

What you end up with. A list of DM conversations still waiting on you, with what was asked and roughly when.

Start with it bringing you the information and you deciding.

Never without. Keep the output of this to yourself. These are private messages from named individuals, and a DM digest is unusually easy to paste into a shared channel without thinking about it.

Built on: Get recent direct message events / Get DM events for a DM conversation / Retrieve DM conversation events

Lists are writable, so the assistant can create one, add accounts by handle, and then read that list's timeline as a filtered feed. The reading part is the payoff: a curated list timeline is a much smaller and more useful thing to summarize each morning than the algorithmic home feed. It can also read who is already on a list before adding to it.

What you end up with. A working list of the accounts that matter to you, and a daily read of just their posts.

Start with it doing the work and checking with you.

Never without. Set the list to private when you create it, unless you have decided otherwise. A public list on X is visible to the people on it, and a list called competitors is a thing you would rather not have named accounts discover.

Built on: Create a list / Add a list member / Fetch list members by id / List posts timeline by list ID

Posting is available, including quote posts, replies, polls and posts with images. Images upload through a separate action and the returned media reference expires, so the upload has to happen immediately before the post. The workable arrangement is the assistant drafting and you approving each one, because there is no scheduling in this connector and no draft state to park things in.

What you end up with. Posts written and published to your account, each one seen by you before it went out.

Start with it doing the work and checking with you.

Never without. Read every post before it goes up. A post is public the instant it is created, there is no scheduled or draft state to catch it in, and deletion does not undo a screenshot.

Built on: Create a post / Upload Media / Get authenticated user

The assistant can find the replies and quotes under your posts, read them, and draft answers as replies to the specific post. Where a reply to a conversation you started needs to go away, hiding it is available and it is reversible, which makes it the sane first move against something abusive. Hiding only works on replies to posts you authored.

What you end up with. Drafted replies to the conversations under your posts, and the abusive ones hidden rather than argued with.

Start with it doing the work and checking with you.

Never without. Approve each reply yourself and keep anything contested out of scope. A reply from your account is public, permanent in practice, and read as the company's position whether or not you meant it that way.

Built on: Search recent tweets / Look up post by id / Create a post / Set reply visibility

Deleting a post is available and your assistant can find the post first so you are not hunting for an ID. It is worth being clear-eyed about what deletion does: the connector's own note is that a deleted post may still sit in third-party caches, and anyone who saw it still saw it. Delete is the right move to stop further spread, not a way to make something not have happened.

What you end up with. The post removed from your timeline, with a record of what it said in case you need it later.

Start with it doing the work and checking with you.

Never without. Name the exact post ID and keep a copy of the text first. Deletion on X is irreversible, it takes the replies underneath with it, and the post can persist in third-party caches and other people's screenshots regardless.

Built on: Search recent tweets / Look up post by id / Delete tweet