Copilot CLI vs CC capability difference

[richlander]

Was working with Copilot CLI this morning, working with llms.txt format.

I asked it to look at this file: https://code.claude.com/docs/llms.txt. And then regroup the content based on the human-provided TOC view at https://code.claude.com/docs. It wasn't able to do it:

The HTML is too large. Let me look at the image you mentioned - can you
describe what the actual section headers are from the screenshot, or should
I try to extract the navigation structure differently?

Can you tell me what the actual section headers are from the screenshot you
referenced (https://github.com/user-attachments/assets/4705c160-7b61-4db1-8
a80-a94d9837c26c)? That way I can recreate the llms.txt with the correct
categories that Claude Code actually uses

I asked Claude Code to do the same thing, asking if it could see the TOC.

It responded with the following and then completed the request.

I can see it.

$ npm list -g
/home/rich/.nvm/versions/node/v24.4.1/lib
├── @anthropic-ai/claude-code@2.1.2
├── @github/copilot@0.0.377
├── @openai/codex@0.79.0
├── corepack@0.34.5
├── markdownlint-cli@0.47.0
└── npm@11.7.0
$ head /etc/os-release -n 1
PRETTY_NAME="Ubuntu 24.04.3 LTS"