How inksss Reviews HTML Projects Before and After Publishing

When you click Publish, inksss first syncs and freezes the current draft, page structure and linked resources into a publication revision. The publication service does not expose the browser upload byte-for-byte; it runs the revision through server-side security processing first.

That processing removes dangerous elements, event attributes, automatic redirects, unsafe protocols such as javascript:, unsafe links and unsafe styles. It also checks the HTML file size and the total page-resource size. Any changes made during processing are shown in the publication result.

To reduce propagation and execution risk while a page is being reviewed, public responses for pages under review temporarily remove scripts and enable a script-blocking policy. Interactions, dynamic data or animations on the page may therefore be unavailable for a while.

After the review passes, scripts may be retained when the current site policy allows them, but they do not run in the inksss main-site context. They run inside an isolated iframe instead. When the site policy disables scripts, they are removed during publication security processing.

Script isolation is a risk-reduction measure, not a guarantee that code is legal or absolutely safe. Do not upload code that steals information, redirects visitors deceptively, imitates a login page or makes malicious requests.

After security processing, the system checks visible and extractable page text. This includes body text, titles, image alt, title and aria-label attributes, visible content text in styles, and the domains used by external links.

inksss maintains screening rules organized by risk category. A match against a blocking rule stops the revision from passing. A match that needs further review pauses publication or sends it into the next review stage. The interface shows a risk category and guidance, but does not expose the exact matched word or internal rule details.

After the synchronous rule checks pass, inksss creates an asynchronous AI review task to analyze the sanitized, frozen HTML for that publication revision. A review-service failure is not treated as a pass.

When the AI review passes, the page continues through publication and its review limits can be lifted. If the AI identifies a clear risk, the revision is rejected. If further confirmation is needed, it enters human review. Public access may remain temporarily limited while the review is in progress.

If the project has an earlier eligible version, a risky new revision may fall back to that earlier available version after the review result is applied. You can update the content and publish a new revision based on the guidance shown in the product.

Rule matches that need confirmation, AI review results or reports submitted on public pages may enter the staff review queue. Staff inspect the page revision, review records and relevant risk information in an isolated preview environment rather than running the pending page in the inksss main-site context.

Depending on the result, staff may restore display, keep access restricted or stop displaying the page. The author is notified through in-product notifications. If a revision does not pass, follow the guidance, update it and publish again. Public pages also have a report entry point; submitting a report sends it to the handling queue but does not automatically take the page down.

After publishing, a page may be under review, access-limited, published, rejected or offline. Pages under review have restricted public access. Normal publication access is provided after the review passes; a rejected revision is not kept publicly available as that revision.

Security processing, keyword screening, AI review and human review are risk-control measures. They are not a warranty that page content is lawful, accurate, truthful or free from third-party rights issues. You remain responsible for the text, images, code and external links in your HTML, and for having the rights needed to use those materials.

Do not use inksss to create or distribute illegal or non-compliant content, fraud or traffic diversion, infringing material, malicious code or other content that harms people. For your detailed obligations and our response measures, also read the Terms of Service.