A Client Content Review Packet for Claims, Visuals, and Reporting

A client approves the article's message, then someone notices that its headline statistic comes from an old screenshot. The banner looks finished, but it implies a result the article never demonstrates. Meanwhile, the performance report says the campaign is working because impressions rose, without saying which posts or replies were counted.

These are different failures with a common cause: the agency has no explicit record of what is ready, what is verified, and what a number actually measures. A final content check should make those boundaries visible before publication, then preserve them in the report afterward.

Agentic Workers provides separate client workspaces and an approval-oriented setup for agent work. The following is a proposed Super Agent review workflow for a service provider. It is not a tested X analytics connection, a guarantee of factual accuracy, or permission for autonomous publishing.

Start with a release packet

Collect the exact draft, approved audience and offer, source links, planned visual, intended destination, and the person authorized to approve publication. Use one identifiable version so a reviewer is not checking a paragraph that has already changed.

Keep the packet inside the client's workspace. Remove customer information that is unnecessary for checking the claims. Do not mix analytics exports, unpublished campaigns, or visual references from different clients, even when the agency serves similar businesses.

The main agent can coordinate the review assignment and outstanding questions. Give the review task permission to inspect supplied material, not to alter public posts or publish replacements. If saving review notes requires a write to a shared tool, define and approve that action separately.

Check claims against evidence, not confidence

Ask the Super Agent to extract the article's checkable claims before judging the prose. A useful record includes the exact claim, supporting URL or document, source date, relevant passage, and any important limitation.

A statistic needs more than a matching number. Check the population, measurement period, definition, and whether the source supports the wording. A survey about respondents' expectations does not establish actual business results. A vendor's example does not establish what every customer will experience.

Require a clear outcome for each claim: supported as written, needs narrower wording, needs a better source, or remove. An inaccessible source belongs in the unresolved category. The agent must not fill that gap with a plausible citation or a confident paraphrase of a search snippet.

Separate client statements from independently verified facts. If the draft says a client achieved a result, require documented permission and evidence for that exact claim. Ordinary editorial approval does not authorize publishing customer-specific performance information.

Brief the visual from the approved message

Write the image brief after the claims review. Describe what the picture should help the reader understand, the approved brand assets, the format, and what it must not imply.

For a guide about preparing a monthly report, the visual might show information becoming easier to review. It should not depict an invented product dashboard or a chart suggesting measured gains. Use original or properly licensed references and do not copy another brand's design merely because it appears in the research packet.

Treat generated text, logos, interface elements, and numbers as things to inspect rather than things to trust. Check the finished export at its actual display size. Confirm that it does not crop essential content or contradict the article. A good prompt is an instruction, not proof that the resulting image passed review.

The reviewer should be able to say which exact article and image are approved together. A later edit to either one should reopen the relevant checks rather than inheriting an approval for different material.

Define the report before collecting results

A performance summary needs a declared scope. Record the account, reporting timezone, date range, included post IDs, and whether replies, paid distribution, and deleted posts are represented. Use an authorized account export or a verified permitted connection. Do not assume the product can retrieve native X analytics simply because agents can connect to other tools.

Keep missing values distinct from zero. If reply impressions are absent from the export, say they are excluded. If two exports use different definitions, do not combine them into a trend without explaining the difference.

Have the Super Agent produce observations separately from recommendations. "This supplied export contains more impressions for these posts" is an observation. "The opening caused the increase" is a causal claim the export usually cannot establish on its own.

Choose the next editorial test from the client's goal. If the goal is useful enquiries, inspect the available enquiry evidence rather than treating reach as a substitute. When that evidence is unavailable, report the limitation plainly.

Make the final decision easy to inspect

Return one release note with four sections: claim issues, visual issues, unresolved permissions, and the requested decision. Include links to the exact assets. Publishing remains with an authorized person or an explicitly approved, bounded release process.

After release, keep the final version linked to its reporting record. If a correction is needed, document the affected claim and approved change. Do not silently rewrite history to make the original review appear successful.

Trial this process on a single article before expanding it. Compare every extracted claim with the draft, inspect every citation, and check the report against the supplied export. Measure missed issues and unnecessary flags as well as completion. Stop when the workflow cannot identify the correct version, lacks source access, or receives material from the wrong client.

This design draws on Ryan's content-agent ideas, especially fact checking, analytics review, and visual prompting. It combines them into a client content review packet without adopting the source's growth claims.

Review Agentic Workers connections before choosing your client workflow's data access.