dsh-interview-forge

Ad-hoc technical drills usually fall apart the moment the questions are asked: there is no structured answer flow, no root-cause attribution, and no way to compare sessions over time. dsh-interview-forge stitches question generation, an in-browser answer overlay, three-stage attribution analysis, and a radar report into a single persistent loop that lives alongside DSH and survives process restarts.

Say something like "start a drill" to the agent and it decomposes the target topic into five to twelve questions. A lightning-bolt overlay slides in at the corner of the browser window with a live timer; multiple-choice items require a written rationale. Hitting "finish" triggers a three-pass attribution over cognitive tags, cross-validation, and narrative risk, then renders a radar chart you can read right in the overlay. A monthly calendar view lets you scroll back through every session, and a bundled skill keeps the agent aligned on question quality and attribution rigor with no extra setup.

This plugin is built for developers running a DSH web profile who want repeated, structured drills in a specific stack and comparable feedback after each round, rather than another pile of unreviewed Q&A logs.