Shell commands like cat trigger Copilot CLI "thinking" instead of executing

[jasper2virtual]

Describe the feature or problem you'd like to solve

No response

Proposed solution

Description:

I'm using GitHub Copilot CLI in the terminal, and I expect it to behave like a shell-native assistant. However, when I type simple shell commands like cat filename.py, Copilot CLI often misinterprets them as natural language prompts and starts "thinking" instead of executing the command.

This breaks the terminal-native workflow and forces me to either:

• Prefix commands with --shell (which is clunky for frequent use)
• Open a separate terminal just to run basic shell commands

Expected behavior:
Copilot CLI should:

• Automatically recognize common shell commands like cat, ls, pwd, etc.
• Execute them directly without triggering AI inference
• Provide a fallback or toggle to switch between shell and Copilot modes more fluidly

Why this matters:
Copilot CLI is designed for terminal-first workflows. Having to jump between contexts or prefix every shell command undermines its usability and purpose.

---

Let me know if you want to add more details or frustrations — I’ll help you polish it further.

Example prompts or workflows

No response

Additional context

No response

[williammartin]

Hey, you can prefix your commands with ! to have them run as shell commands:

I don't think we'll be changing the core loop to recognise commands because it just flips the problem. Sometimes people paste in commands just to get information on them. We'd rather avoid a heuristic here and just require !.

Prefix commands with --shell (which is clunky for frequent use)

Don't think this is a feature but I assume this was generated by an LLM.

Thanks!