Do numbered marks + spoken notes beat one big screenshot for coding agents?

When I paste a screenshot into Claude Code or Cursor, a crop plus an arrow usually fixes the "where." What kept going wrong for me was the "why" once there are three or four things to fix on one screen.

What helped was talking while I draw. I circle a clipped label and say "let this wrap," then draw an arrow to a chart point and say "tooltip on hover and keyboard focus." Lasso puts [1] and [2] in the transcript at the exact words I said while making each mark. So the agent gets numbered marks with my intent attached, instead of a paragraph it has to match back to the image.

I also stopped pasting. A local coding agent on the same Mac can pull the latest annotated capture or voice brief itself through Lasso's local MCP server, and Lasso asks before the first agent call.

I wrote up the full workflow, with an example prompt for a UI fix, here:

Curious how other engineers handle this. Do numbered marks cut down on wrong-element edits for you, or do you still describe everything in text?

8 views

Add a comment

Replies

Be the first to reply

Have a question or a thought to share? Add a comment above to start the conversation.