A brand-governed content repurposing agent turns a bounded set of social references into original, reviewable assets and copy. Growthub's walkthrough starts with a selected profile, asks the agent to inspect engaging posts, brings in approved Brand Kit references, and saves the resulting assets and captions to Google Drive and a Google Sheet. The critical handoff is a complete review record that connects each output to its inspiration and intended account.
TL;DR
Choose the source profile and limit the research task before generating anything. Inspect the selected posts and preserve the observed engagement context. Use the Brand Kit to guide the new content, then review the actual images and copy. Save the asset, caption, target account, and inspiration together. Generation is one stage; approval and scheduling follow as separate decisions.
Begin with a bounded research question
At 0:08, Antonio describes pointing the agent at a profile. The walkthrough names Instagram, X, and LinkedIn as places where the agent can inspect posts. Its purpose is to identify useful creative inspiration and turn it into content for the intended brand, rather than treating every post as an equally suitable source.
The example becomes more concrete at 1:09. Research is given a duration limit, and the agent is asked to identify three static posts. This is a valuable constraint: it defines what the researcher should return and prevents an open-ended browsing task from becoming the whole project.
In your own brief, name the profile, the audience you want to reach, and the kinds of posts worth studying. Ask the agent to preserve the source links and the engagement observations it actually sees. The goal is a small collection of inspectable references that can support a clear creative decision, not a large folder of disconnected screenshots.
Extract a useful idea before making a new asset
The source describes reverse-engineering engaging posts and using them as inspiration. A practical way to apply that idea is to separate a reference's communication structure from its exact execution. Identify the question it raises, the way it organizes the answer, and the reason the format may fit your audience.
Then write a new brief for your brand. A reference might suggest a comparison, a short explanation, or a visual joke, but the resulting content still needs its own message and relevant brand context. This is an editorial recommendation for applying the demonstrated inspiration workflow; it is not permission to copy another creator's expression or imply that their audience results will transfer.
Keep selection evidence beside the brief. Engagement observations tell the reviewer why the post attracted attention during research. They do not guarantee that a new version will perform similarly. Preserving that distinction keeps the research useful without turning a creative hypothesis into an unsupported performance claim.
Bind generation to the Brand Kit
At 1:29, the walkthrough combines an image-generation skill with the Brand Kit API skill. Antonio explains that the agent retrieves references such as the Growthub logo and visual guidance before producing the new content. The source also describes using an authorized personal likeness in the demonstrated examples.
For a new run, bind generation to the references approved for that exact brand and assignment. The relevant question is not whether the model can draw a plausible logo or person. It is whether the resulting asset follows the actual references and the task's intended use. If a likeness is not part of the authorized brief, it should not be added simply because the demonstration includes one.
The first part of the demonstrated run checks plugin and Brand Kit access. That check comes before the finished outputs. Adopt the same ordering: establish usable source and brand access, complete the creative brief, and then generate. It is much easier to assess a result when the target treatment was specified before the image existed.
Treat the finished image and caption as one item
Around 2:19, the walkthrough shows generated examples. It describes six variations associated with three static references, including personal and Growthub-branded treatments. That is a particular example of a production set, not a required count for every future run or a guarantee of how long production takes.
For each output, inspect the image alongside its caption. The visual should communicate the same idea as the copy, and both should suit the intended account. Review logo treatment, readable text, factual language, and whether the content offers a useful takeaway to the audience. These are recommended editorial checks for making the source's branded-output step reviewable.
Variations should also have a reason to exist. Different account identities may justify different treatments, as shown in the source. A different explanation or visual structure may justify another version. Repeating the same idea with a small cosmetic change can create more files without making the final content set more useful.
Save the context needed for approval
At 2:44, Antonio walks through the Sheet fields: post ID, asset link saved in Drive, caption, target handle, original inspiration, and relevant engagement statistics. This is one of the most actionable parts of the demonstration because it defines a handoff that another person can inspect without reconstructing the whole chat.
The asset link gives access to the deliverable. The caption shows the proposed message. The target handle identifies the destination. The inspiration and observed engagement preserve the reason for selecting the idea. Together, these fields connect research, production, and review while keeping the output anchored to a specific publishing context.
For an implementation, verify that the links open and the saved caption is the intended version. A generated asset that was never saved, or a Sheet row that points to the wrong version, is an incomplete handoff. The source's workflow ends with material that can be approved and scheduled; the existence of the row should not be mistaken for evidence that either later action occurred.
Turn the instruction into a recurring workflow carefully
At 0:54, the video explains that the instructions can be used in a routine or scheduled task. Reuse the method while keeping the run's source profile, research limit, brand references, and destination explicit. A recurring instruction is useful when it preserves those boundaries rather than silently expanding the source pool or publication scope.
The review record is the bridge between runs. It lets a team see what was already researched and produced, compare outputs, and decide which ideas deserve another iteration. This article recommends retaining that context as part of the workflow. It does not claim that the source demonstrates every possible duplicate check or scheduling integration.
Expert Q&A
What should the research agent return before image generation?
It should return the selected references and the observed context needed to justify them. In the example, the task is bounded to three static posts and includes engagement inspection. A clear selection brief gives the generation stage a specific idea to interpret.
Why use the Brand Kit instead of describing the brand from memory?
The demonstration uses the Brand Kit API to retrieve the references for the run. That gives generation an explicit visual basis, including the approved logo. The asset should be judged against those actual references rather than an improvised description of the brand.
What belongs in the review Sheet?
The source lists a post ID, Drive asset link, caption, target handle, original inspiration, and engagement statistics. Those fields make the output traceable and give a reviewer the material needed to assess the proposed post in its intended context.
Does creating the review record publish the content?
No. The walkthrough describes saving material so it can be approved and scheduled. Treat the generated output, saved review record, approval decision, and scheduled post as separate states. Each should be verified in the system responsible for that action.
Make the next creative handoff complete
Start with a small reference set and finish with assets, copy, and source context that a reviewer can actually use. Explore Growthub to discuss brand-governed agent workflows for your content operation. The immediate goal is a clear, inspectable production process that can support a deliberate publishing decision.
Source: How to Build a Social Content Repurposing Agent (Full Walkthrough), Antonio Romero, September 17, 2026. This article adapts the walkthrough; implementation recommendations are identified as such. Automatic captions were reviewed, and ambiguous wording was excluded.