Content production, from idea to published
Run research, drafting, review and publishing as one workflow: agents do the legwork, humans hold the gates that matter.
Opens in AGNT and takes about a minute.
Didn’t open? AGNT may not be installed here yet. Download AGNT free · Download the .json
Content production, in brief
| Tools it needs | Notion, Google Drive, X (Twitter), YouTube, Unsplash |
|---|---|
| Setup | 4 steps, about ten minutes |
| Approval | Yours, per category — nothing is sent on your behalf unless you say so |
| Where it runs | Your own machine. Credentials stay in a local vault. |
What the agent actually does
A first draft that is already sourced
Before anything is drafted, an agent gathers and reconciles sources, so the piece is built on material somebody can check rather than on the model’s recollection of the internet.
Draft in your voice
Style lives in a skill the agent loads — your vocabulary, your structural habits, the things you never say. That is a reviewable file rather than a prompt somebody tweaks until it sounds acceptable.
Publish everywhere at once
One approved piece becomes the post, the thread, the newsletter section and the summary, each shaped for its destination rather than the same text pasted four times.
The tools this job needs
Notion
Where the work arrives. The team wiki agents can actually keep current. The agent watches it and reads what turns up in full, rather than matching a rule against a subject line.
Google Drive
Context the agent pulls in before deciding. Where the documents live — and pile up. It is read, not just referenced — which is what lets the decision account for it.
X (Twitter)
Context the agent pulls in before deciding. The feed where your market thinks out loud. It is read, not just referenced — which is what lets the decision account for it.
YouTube
Where the result lands. Long-form content worth repurposing everywhere else. Nothing is written here until the agent has formed a view and, where you asked for it, you have approved it.
Set it up
- Capture your voice and standards as a skill the drafting agent loads.
- Set the research step to gather sources and produce an outline for approval.
- Draft, self-edit against the standard, and route to a human for the publish decision.
- Fan the approved piece out to each channel in that channel’s native shape.
You can see exactly what it did
The brief it works from
This is the actual instruction set the template installs — what the agent is told to do, and what it is told never to do. Every line of it is yours to edit in AGNT after install.
Read the full brief
You run content production from research to a draft a human approves.
Work in stages, not in one leap
Research, outline, draft, self-edit, adapt per channel. Asking for a finished article in a single pass produces something plausible and unusable, because writing is not one task. Each stage has different inputs and different ways of going wrong.
Research first, always
Gather and reconcile real sources before drafting anything. Every factual claim in the finished piece must trace to something you actually read, not to a recollection that sounds right. Technical claims trace to documentation or source code. If you cannot source a claim, cut it — a piece that is 10% shorter and entirely defensible is strictly better.
The outline is a checkpoint
Produce it and stop. A human approves the angle before you write, because a beautifully written article about the wrong thing is a total loss and it is cheap to prevent.
Voice
Load the user's style guide and follow it specifically: their vocabulary, sentence rhythm, structural habits, and the constructions they never use. If no style guide exists, ask for examples of their best writing before drafting, and say plainly that without it the output will read generic. That is the honest answer and it is more useful than producing something bland and hoping.
Avoid the tells: opening with a rhetorical question, "in today's fast-paced world", triads of adjectives, summarising what you just said, and closing paragraphs that restate the introduction.
PER CHANNEL, REWRITE — DO NOT REFORMAT A thread needs a first line that earns the second. A short post needs the whole idea to land at once. A newsletter can take a longer run-up. The same text pushed everywhere is optimised for nowhere.
You never publish. A human approves every piece before it goes public.
Shown exactly as it ships. The agent inherits whichever model you already use.
Pipelines beat prompts
Asking a model for a finished article produces something plausible and unusable, because writing is not one task. Research, outline, draft, edit against a standard, adapt per channel — each is a separate step with its own inputs and its own failure modes. Modelling them as separate steps means you can put a human gate exactly where taste is required and automate the rest honestly.
Voice is the part that has to be captured, not prompted
The reason machine-written content reads like machine-written content is that nobody encoded what makes their writing theirs. Doing it properly means writing down the specifics: sentence rhythm, the constructions you avoid, how much hedging is acceptable, worked examples of good and bad. Kept as a skill, that becomes an asset the whole team shares rather than a knack one person has.
The gates that actually matter
Two are non-negotiable: a human approves the outline before drafting, and a human approves the piece before publishing. The first prevents a beautifully written article about the wrong thing; the second is the last line between your brand and an embarrassing sentence. Everything in between can run unattended.
Common questions
Will this produce content that reads as AI-written?
It will if you skip the voice work. A captured style guide with real examples is the difference, and it is the step most teams skip and then blame the model for.
Can it publish without me?
Technically yes; advisedly no. The recommended shape keeps a human approval before anything goes public.
Does it work for technical content?
Yes, and the citation step matters more there. Technical claims should be traceable to documentation or source, which is exactly what the research step produces.