Portable participant instrument

Prepare one HTML file that works without a study account.

Short answer: freeze an edition, download its participant HTML, test that exact file in the target browsers, give the participant an approved session code and return route, then import the resulting integrity-checked JSON response.

Before you freeze

Replace default participant-information text with reviewed study-specific wording.
Confirm no card label, explanation, or category reveals confidential information beyond the approved study scope.
Check the study method, minimum/maximum groups, all-card rule, category-name rule, final comment, shuffling, and category capacities.
Use researcher notes only for private material; they are excluded from the participant edition.
Confirm the participant has a safe place and approved method to return the JSON file.

Pilot the exact file

Open the generated HTML directly with file:// in every target browser.
Complete a session using only keyboard controls: select a card, move it to a group, and reorder it.
Repeat at 320 CSS pixels or a small phone viewport and at 200% zoom.
Try the longest labels, the maximum intended groups, and every category capacity.
Download a saved partial response, close the page, restore it, and continue.
Complete and download the final response. Verify that no network request was made.
Import saved then completed revisions into a disposable copy of the researcher project.

Issue package

Provide the participant with the exact HTML file, the approved session code, a short instruction for opening downloaded files on their device, the approved return route, and a researcher contact outside the tool. Do not use their name, email, phone number, employee number, or another direct identifier as the session code.

Save and resume behavior

The participant file deliberately does not use browser storage. “Save progress” downloads a saved JSON file. Restoring it verifies the edition identity and SHA-256 digest, restores the exact group/card order, and increments the next response revision. This makes recovery portable and visible but requires the participant to keep the file.

Completion behavior

The finish check lists unmet rules. When all rules pass, the page downloads a completed response. It does not upload, email, or submit it. The participant must return it using the researcher-provided method.

Operational warning: a standalone file reduces platform dependency, not human handling risk. File names, transfer channels, endpoint backups, shared folders, device encryption, access lists, and retention/deletion all need an approved process.