X posts agent
How the X posts agent listens before it writes, what it learns from your own posts, and how an approved post reaches your account.
What it does
The X posts agent keeps your own posts planned a week ahead. It reads the Calendar first, what is already scheduled or waiting on X and on LinkedIn, then listens before it writes: a handful of searches for what your audience is talking about, then a post in your brand voice for each coming day that has none on X, each on its own day and hour.
The week is spread across your product's three to five content pillars, the topics it owns, each post on one of them, with promotion kept to the small share the social skill gives it. On the skill's thread days the post is a thread, built on the key takeaways of one of your published articles when the Calendar lists one. Every draft follows the agent's X playbook: the hooks, formats and calls to action that work for the high-engagement accounts of your niche.
Links stay out of the body — they belong in a first reply. A post that quotes somebody else's post is built from what the search returned, never from text the model retyped, so what you read on the proposal is what X served.
When it runs
It runs as a Loop, on from the moment the connection exists, since nothing is posted without your approval. Each run fills the days of the coming week the Calendar leaves empty, so the channel still posts every day while the agent runs only every few days; a run that finds the week covered ends with no proposal and says through which day it is covered. The cadence comes from the agent, shows on its page and is not editable, and it carries no weekly cap: rejecting the posts you do not want is the throttle.
What it needs
The X connector. Without it the agent is unavailable and does not run. Replying to other people's posts is a separate agent, X replies, on its own switch.
What it reads
The foundation doc, the growth plan and the product memory first. Then X.
- Your own posts and what they did, synced into a connector snapshot at the start of the run, so the agent writes against its own results rather than in the dark.
- Its X playbook, kept as a data output. When it is missing, or your best posts contradict it, the agent rebuilds it before drafting from the top posts of 10 to 20 high-engagement accounts of your niche, the accounts the X replies agent engages with first, and saves it again.
- The day's search, up to five queries for this agent: the threads worth learning from, the angles a post could take. Kwern reads through its search service and falls back to the official API only when that refuses, and a day where X cannot be read at all is written from the product alone.
What it leaves you
Proposals in the post category — your own post, or a quote of a post one of the searches returned. X is a channel Kwern can write to, so a proposal ends on approval: you edit the text if you want to, and the post arrives planned for a day and an hour, so Approve for Tue 09:00 agrees to that hour and Kwern posts it from your account then and links you to it. Publish, in the schedule menu beside the button, sends it at once. A publish that fails can be retried, or marked as done if you posted it yourself in the meantime. A quote has no approve: X's API no longer takes quotes, so you copy the text, open the quoted post from its link, quote it from your account and mark the proposal as done.
The skills it runs
It enters through jab-hook, and may activate social and copywriting.
Editing its instructions
The agent's page in the app carries the instructions it is given at the start of every run, and you rewrite them there — what to listen for, how long a post should be, the subjects that are yours and the ones that are not. The cadence is not editable. The switch reads on, off or paused, where paused means Kwern stopped it for a reason and will resume it when that reason clears.