ORIGINAL RESEARCH / BROWSER HANDOFF TEST
Does a filtered review keep its complete export?
We hid seventeen checks on screen, downloaded both formats, and inspected what survived.
A filtered view can help a reviewer concentrate on open work. It must not silently turn the downloaded handoff into a filtered, incomplete record. We tested the public 18-check release-review tool with fictional entries so the result is reproducible without anyone's product or client data.
This is a test of this tool in one Chrome session. The fictional “pass” and “follow-up” labels below are sample data; they are not verdicts on an actual release.
What the browser run showed
- With the “Needs follow-up” filter selected, the page showed one check. The summary still showed one pass, one follow-up, one not-applicable item, and fifteen unchecked items.
- The saved Markdown and JSON files each contained all eighteen checks and all three fictional notes. We verified the actual downloaded files, not merely the button message.
- After a later unexported edit, an in-site Home link navigated away without a captured warning; returning showed a blank review. A full reload emitted a browser
beforeunloadevent, but automation handled that dialog, so this run cannot say what a person saw.
What changed after the test
The baseline exposed a missing warning on internal site links. We added a native page dialog for edits that are not in the last export. In a separately built local preview, clicking Home opened the dialog; “Keep editing” retained the input; “Leave and discard review” navigated away; and a blank review left without prompting. The five follow-up observations preserve this correction separately from the live baseline.
This does not automatically save the review or cover every way to leave a tab. An export request clears the warning state before a saved file is proven, so the tool still asks readers to check their Downloads.
Method and test conditions
We used Chrome 153 on macOS 26.6.2 at a 2560 × 1353 CSS-pixel viewport and device-pixel ratio 2. We entered one fictional follow-up, one fictional pass, one not-applicable reason, and left fifteen checks untouched. The primary export run started on September 18, 2026 at 16:05 UTC.
We then filtered to follow-up, requested both downloads, inspected the saved files, edited again, and navigated away. Read the complete method and provenance.
Recorded steps
| ID | Action | Observed result |
|---|---|---|
| H01 | Open public release-review tool in a new tab | 18 not checked, zero in the other three categories |
| H02 | Enter fictional conditions and three synthetic statuses with notes | Browser summary showed 1, 1, 1, and 15 respectively; all notes appeared on their rows |
| H03 | Select Needs follow-up filter | One review row visible and status text said 1 check shown; summary remained 1 pass, 1 follow-up, 15 unchecked, 1 not applicable |
| H04 | Use Export Markdown while filtered | Browser UI said download requested; 8,373-byte Markdown file appeared on disk at 16:06:23 UTC and offline verification found 18 check headings and all three notes. Browser waitForEvent(download) timed out, so file existence and content were verified directly. |
| H05 | Use Export JSON while filtered | Browser UI said download requested; 10,054-byte JSON file appeared on disk at 16:06:50 UTC. Offline verification found 18 unique IDs and all 18 statuses, including hidden pass, N/A, and 15 unchecked. |
| H06 | After export, edit the fictional release field and click the site's home link, then revisit the tool | Navigation completed to the homepage without a captured JavaScript dialog event. Returning to the tool showed 18 unchecked and blank conditions; the new edit was not retained. |
| H07 | Edit a fictional field and perform a full-page reload | CDP captured one Page.javascriptDialogOpening event of type beforeunload for the tool URL. The automation did not expose a visible prompt and the reload completed to a blank review, so this does not prove what a human user would see or choose. |
Representative screenshots
Inspect the handoff files
The sample files are complete tool outputs from the fictional run. Compare their check IDs, statuses, and notes with the filtered screenshot.
Limits and corrections
- One browser, one operating system, one viewport, and one primary export run cannot establish cross-browser behavior, accessibility, security, or production readiness.
- We did not capture a network trace. The screenshot and exported files do not prove what a server did or did not receive.
- The browser automation handled the full-reload dialog; we recorded the event, not a verified human-facing prompt.
- This baseline record documents the navigation problem as observed. A correction requires a separately verified follow-up run rather than rewriting the original observation.
Version 2026-09-18.1 preserves the original seven observations and adds five separately dated local correction checks. The original navigation finding is unchanged.
Cite and reuse this record
Awesome Patel (2026). “Does a filtered review keep its complete export?” SaaS Mirror, browser handoff study, version 2026-09-18.1. https://saasmirror.com/research/release-review-handoff-study
The JSON and CSV observation data are dedicated to the public domain under CC0 1.0. Citation is appreciated, not required. The screenshots, article text, and sample tool exports follow the site terms.