SMF·TANGO
Can a prompt replace Tango?
Screen recording — screen recording, walkthroughs and product demos
Exhibit tracking slip
Verdict
The step-guide half of Tango is already covered elsewhere in this catalogue. What distinguishes Tango is the live walkthrough: replaying a captured sequence as an overlay on the actual application, highlighting the next element to click as you go. That needs a stable way to find an element again on a page that has changed since capture, which is a genuinely awkward problem and the reason this build takes a weekend rather than an evening.
Exhibit A — The prompt
Received on31.07.2026Build a browser extension that captures a workflow and can replay it as a live overlay on the real application.
Capture: on record, listen for clicks, form input and navigation. For each step store: the page URL pattern, a screenshot of the viewport with the target element outlined, the text label of the element, and — critically — a selector strategy rather than a single selector.
The selector strategy is the technical core of this build. A single CSS path breaks the moment the application ships a change. Instead, record several candidate locators per step in priority order:
1. A stable test attribute if the application provides one.
2. Accessible role plus accessible name, which survives most visual redesigns.
3. Visible text content, exact then normalised.
4. A structural CSS path, as a last resort.
At replay, try each in order and use the first that matches exactly one element. If several match, or none does, stop and tell the user which step failed and what it was looking for, rather than clicking the wrong thing. Guided software that clicks the wrong button is worse than no guidance.
Redaction at capture time, not after: before storing a screenshot, blur the contents of any password field, any input whose value matches a card-number or long-digit pattern, and any element the user has added to a per-domain blocklist. Redaction has to happen before the image is written, because a screenshot blurred afterwards has already existed unblurred on disk.
Two outputs from one capture:
- A static guide: numbered steps with the annotated screenshot and an auto-written instruction from the element label and action, all editable.
- A live walkthrough: overlay mode on the real application, highlighting the located element for the current step with a tooltip of the instruction, advancing when the user performs the action rather than on a timer. A progress indicator, a skip, and a stop that leaves no trace on the page.
Overlay implementation: render into a shadow root attached to the document so the host application's CSS cannot interfere and the overlay's styles cannot leak. Position by the located element's bounding box, updating on scroll and resize.
Storage: guides in extension storage with a JSON export and import, so a guide can be shared as a file.
Out of scope: capture outside the browser, hosted sharing or a guide library, viewer analytics, video recording, and any automatic clicking — the overlay guides a person, it never performs the action for them.Opening prefills the prompt — press enter to run it.
Exhibit B — What you lose
- B.1 capture outside the browser — desktop applications are out of reach
- B.2 team libraries with shared guides and permissions
- B.3 hosted sharing with a link and viewer analytics
- B.4 the blur controls maintained against real customer data patterns
Exhibit C — Why people still pay: capture polish, hosting, and collaboration
Because guides are made to be shared with a team, and a hosted library with permissions, links and analytics is the half of the product a browser extension on your own machine cannot supply.
Questions
Can I import my Tango guides?
Tango exports guides as PDF and to some document tools, which preserves the steps as reading material but not as replayable captures. The selector data is Tango's own, so a live walkthrough has to be re-captured — which takes as long as performing the workflow once.
How often does the overlay break?
Whenever the application changes the element it points at. The layered locator strategy survives most visual redesigns because accessible role and name usually persist, but a renamed button breaks the step. The build fails loudly at that point rather than clicking something else, which is the correct behaviour and still a broken guide.
What does it cost to run?
Nothing. Everything lives in the extension and on local storage, with no server and no account. Sharing means sending a JSON file, which is the trade for having no hosting cost.
What is the one thing that does not survive the rebuild?
Sharing it. Tango's value is a library the whole company can search, with a link you paste into a ticket. A guide exported as a file and emailed around is the same content with none of the distribution, and distribution is what a documentation tool is for.
Related tools
Receipt