TL;DR - feature request: let users reference the chat context in the terminal prompt
ok cursor team, we’re starting to get a lot closer to “cursorbot” (Cursorbot: commit a PR from a GH issue)
here’s my test today: get cursor to write a blog post and commit to git (inside the repo of my blog).
general prompt:
write a new post in “works” about how much i love llamas, create the new file and commit to git. this is a @Hugo website.
test 1: ask the “terminal cursor” to write a new blog post and create the git commit.
result: not so hot.
notes:
- the references available to the “terminal cursor” are not the same as the “chat cursor” (can’t reference folders for example).
- the output is very short
- (unlike interpreter mode) it doesn’t automatically learn from error messages.
example output:
(base) mypath$ echo “—\ntitle: "My Love of Llamas"\ndate: "$(date +‘%Y-%m-%dT%H:%M:%S%z’)"\ndraft: false\nauthor.name: "Raymond Weitekamp"\ndescription: "Expressing my adoration for llamas and their impact on my life."\nimage: "/images/llamas-love.jpg"\n—\n\nEver since I can remember, llamas have been a significant part of my life. Their gentle nature and quirky personalities have always fascinated me. This post explores my journey with these magnificent creatures and how they’ve influenced my perspective on life.\n\n<!–more–>\n” > content/works/my-love-of-llamas.md && git add content/works/my-love-of-llamas.md && git commit -m “Add new post about my love of llamas”
bash: !–more–: event not found
test 2: ask the “interpreter mode cursor” to write a new blog post and create the git commit.
result: better, but can’t go all the way to a git commit
notes:
- i understand the security reason why i need to execute each python script, but it would be more convenient if i didn’t have to click execute each time
- it didn’t get it on the first try, but it was able to see the error message and get it on the second try.
- can’t commit to git, so it leaves instructions instead:
example output:
# Define the content again for clarity
content = f"""---
title: "Llamas: The Love of My Life"
date: "{current_date}"
draft: false
author.name: "Raymond Weitekamp"
description: "An ode to llamas, the majestic creatures that have captured my heart."
image: "/images/llamas-love.jpg"
tags: ["llamas", "personal reflections", "animal love"]
---
Llamas have always held a special place in my heart. From their gentle demeanor to their quirky personalities, there's so much to love about these majestic creatures. In this post, I want to share my admiration for llamas and why they mean so much to me.
Llamas are not just animals; they are companions, protectors, and sometimes, the source of a good laugh. Their curious nature and unique vocalizations make every interaction memorable. Whether it's a trek through the mountains or a quiet day at a farm, being around llamas is always a joy.
Their wool is not only beautiful but also incredibly soft and warm, making it perfect for textiles. But beyond their physical attributes, it's the spirit of llamas that truly captivates me. They are resilient, strong, and possess an unparalleled grace.
This post is a tribute to all the llamas out there and to those who care for them. Let's celebrate these wonderful animals and spread the love they bring into our lives.
<!--more-->
"""
# Add the content to the newly created file
ide.edit(path="content/works/llamas-the-love-of-my-life.md", instruction="Add the content about your love for llamas.")
so here’s my feature request:
let me reference the chat context in the terminal context to finish the job! (and i don’t mean the up arrow to just copy the prompt.)
(& tell me where bash mode is hiding now?)