pinpoint

Debugging a web UI with an AI agent often stumbles on one simple question: which element are you actually pointing at? A vague description like the gray button, third row, second one from the left is hard to verify and wastes turns on back-and-forth clarification. Pinpoint exists to eliminate that friction.

The workflow is straightforward: press Alt+Shift+M on any page to enter mark mode, click the target element, and Pinpoint automatically captures a screenshot, CSS selector, text excerpt, and page coordinates. A comment popup lets you add context, and the moment you send it, the structured card lands directly in your dsh session. The agent receives precise, selector-level context instead of a fuzzy description, so it can pinpoint the issue in a single pass.

Under the hood the plugin has two pieces: a Chrome/Edge browser extension (MV3) that handles marking, commenting, and cross-tab staging, and a dsh plugin that renders the context card and injects the expected format into the conversation. It is built for developers who regularly collaborate with AI agents on front-end debugging, UI walkthroughs, or requirement review—anyone whose counterpart keeps needing to see exactly what you are pointing at. Pinpoint turns a vague you-know-that-thing into a precise, structured card every time.