Browser plugin: add a step that saves the page DOM to a file in the artifacts directory #96

Problem

There is no way to get the page's DOM out of a scenario. I execute the script returns a string into <<script_result>>, but nothing writes a variable to a file, so the only option is printing it to the log, where it is truncated. Writing a selector for an unfamiliar page means a full run per guess.

Proposal

Add two steps, kind action, same shape as I dump the browser console:

• I dump the DOM – writes the serialized current document as HTML into the artifacts directory.
• I dump the DOM of "<selector>" – same for one element's outerHTML (CSS, xpath= or text=, like the other selectors).

Behaviour:

• Never waits (like steps 26–28); use an eventual assertion to arm it.
• The file name does not collide with a previous dump in the same scenario (dom-<n>.html).
• on_failure gains a dom value (off by default); the failure diagnostic carries the file path.

Implementation notes

• Add GET /session/{id}/source to webdriver.rs; the element form uses execute with arguments[0].outerHTML.
• The source endpoint does not include shadow roots or iframe contents. Document this; walking them can be a follow-up.

Open questions

• Masking: hidden fields, tokens and input values end up in the file. Dumps should go through the same masking as other output.
• Screenshots have the same overwrite problem today (screenshot.png).

Note

The steps belong to the browser plugin (bddkit/bddkit-browser, src/steps.rs); filed here for tracking.

Generated by Claude Code