A prompt that takes one idea and turns it into a researched brief, a draft, and the repurposed versions for every channel you use — running on a schedule, so publishing stops depending on whether you got to it that week.
The writing is not the bottleneck. Deciding what to write about, checking what already exists, and finding the three sources worth reading is where every plan quietly stops.
Anyone can name a topic worth writing about. Turning that into something worth reading means finding out what has already been published on it, where the consensus actually is, and which of the obvious points are wrong — and that is a couple of hours of reading that happens silently, which is why it so often does not happen at all.
The work after the draft is mechanical and takes forever. A long post becomes a thread, a thread becomes an email, an email becomes five posts, and each needs a different opening, a different length, and a different level of assumed knowledge. Done by hand it is a full day of reshaping for something that should take minutes, so it is the part that gets cut.
Most content programmes do not fail because nobody had ideas. They fail in a fortnight when the person running it has a bad week, and nothing is scheduled to cover the gap. By the time the rhythm is broken, restarting feels like starting again, so the calendar stays empty and the channel goes quiet for good.
Copy it once, paste it into Zaira, and it takes the calendar from idea to channel: research what is already been said, draft against it, and reshape the result for each audience — with every claim traceable to something it actually read.
Three stages, one goal. The calendar keeps filling itself, so the question stops being what to post and becomes which of the finished pieces is actually any good.
Because the engine runs on a schedule rather than on your initiative, a bad week becomes a gap in review rather than a gap in publishing. Most programmes die from one missed week, and this is the structural fix for that.
The reshaping that used to eat a day per piece happens in one run, with each version opening differently and assuming a different amount of context. That is the difference between repurposing and cutting up a post into fragments.
Every factual claim traces to something the agent actually read, and anything unverified gets flagged rather than smoothed over. It is a meaningful constraint on the failure mode that makes AI-written content unpublishable.
The practical questions people ask before letting an agent plan their content.
No — Step 4 assembles drafts in the tool you nominated and stops there, with nothing scheduled to go live. Publishing is the one step worth keeping as yours, because it is the only point where a human has looked at the piece and decided it is worth your name. Once you trust the drafts after a few weeks, the reshaped channel versions are the safest thing to release first, since they cannot introduce anything the approved draft did not say.
The context section is the answer, and specificity is what counts: two or three real sentences about your voice and house rules, plus an example of something you have written that sounds right. Vague adjectives get averaged into the same neutral prose. Step 1 also helps more than it looks — deciding the angle against what already exists is most of what makes a piece feel written by someone with a position rather than someone summarising.
It replaces the parts of the job that are mechanical — the research sweep, the reshaping, the format changes, the scheduling — which is most of the hours and none of the judgement. A writer using it does less formatting and more editing, and the thing they spend their judgement on is whether the argument is any good. If nobody on the team has that judgement, this produces a high volume of work that does not deserve to be published.
Step 1 is the safeguard: it checks what has already been published on the topic and tells you plainly when a piece would be redundant, proposing a different angle rather than drafting a near-duplicate. That matters most for the long tail, where the same thin article at six URL variants is worse than no article. Do not let the engine publish to a new URL by default — the duplication problem is created at the publishing step, not the drafting one.
The prompt is written for the formats you list in the context section, and each one is a reshaping of the same draft rather than new work, so adding channels is mostly naming. Zapier charges are worth checking against your plan for the volume you plan to run, since that is the part that scales with output rather than with research. Start with two channels and one weekly piece; the value is in the rhythm being unbroken, not in the width.
Only what Step 0 discovers and enables — the actions for your CMS and mailing list, and nothing else. Discovery happens before any action is called, and Step 0 tells it not to guess action names or parameters, because a guessed write action against a real CMS is the failure mode worth designing against. You can revoke the connection from your Zapier account at any time, and keeping it to drafting rather than publishing keeps the blast radius small.