The problem: context
For many of us (including me), a coding agent that can read and edit files in your repository directly is out of reach. Since regular chatbots generally have generous free tiers, that is what I use regularly for coding.
Sometimes I want to give the chatbot some context about my existing code. What I generally do is copy and paste the relevant files, put them in code blocks, and type in the paths to those files.
However, giving the chatbot an idea of how the overall project is structured (files and folders) is crucial for optimal outputs. Doing this manually is really hard.
What I used to do was run the tree command and copy its output. But this includes files that are not included in version control (such as node_modules), which can get high in number and clutter the output. I found out that I can use the --gitignore flag to only include tracked files, but the overall workflow was still slow.
