Cursor 1.6 ships reusable slash commands — team-shareable prompts stored as plain Markdown files in your repo.
The feature works like this: drop a .md file into .cursor/commands/, type / in the Agent input, and pick the command from a dropdown. Because they live in the repo, commands travel with version control. Every teammate who pulls gets the same prompt library.
- Slash commands — reusable prompts stored in
.cursor/commands/[command].md, triggered via/in Agent input - Conversation summarization — long Agent sessions get condensed so context stays useful without blowing up the window
- Improved Agent terminal — terminal interactions inside Agent mode are more reliable
The slash commands feature solves a real friction point. Teams using Cursor’s Agent have been copy-pasting prompt snippets in Slack or Notion docs, then manually feeding them into chat. Now those prompts are code-adjacent artifacts — reviewable in PRs, versioned alongside the codebase, and instantly accessible without leaving the editor.
Conversation summarization matters for longer Agent sessions where context windows fill up. Instead of starting a fresh chat when things get unwieldy, Cursor now compresses earlier turns so the model retains the thread without choking on token limits.
NOTE
The changelog is light on details about the terminal improvements — no specifics on what broke before or what’s fixed. Worth testing if you’ve hit Agent terminal hangs in previous versions.
Slash commands ship in Cursor 1.6 now. Create your first one by adding a Markdown file to .cursor/commands/ and typing / in Agent chat.