Skip to lesson
Exit
Turn a dependency audit into a canvas1 / 2

2 min lesson

Build a dependency-audit canvas

Practice Canvases on one real case. Check the result before moving on.

Step 1 of 2

Ask Agent to run the repository's dependency audit commands and open a canvas for the engineering lead. Require package, current version, target version, advisory, severity and recommendation, sorted by severity. Show the source commands, commit and run time. Open the canvas card and compare its summary and one high-severity row with the terminal output and lockfile. If a number looks stale, ask Agent to rerun the commands and show its work. Add a short compatibility caveat beside affected rows. Before Publish, check the plan, team, admin setting and privacy mode. Open the shared link as the intended teammate. Keep the source commit and run time visible, and point the handoff back to the lockfile and pull request.

Open Canvas, then Keep All the pending files

0:29 · narrated
Read this demo as text
  1. Canvas is its own editor tab — not the in-window Browser. Empty means type slash canvas in chat when you want a real artifact. We are not inventing a filled table here.
  2. Pending file changes still need a decision. Keep All accepts the staged edits; Review walks them file by file — Canvas being empty does not skip that step.
Use this reconstruction for the Canvas tab and file-review boundary. Current docs say to ask Agent for a canvas or run Open Canvas.

Practice next: Ask Agent for one dependency-audit canvas, then check its source data before you publish it.

Simulated Cursor 3.12 (macOS, light) — beta educational reconstruction, not the real product.