Feat: Support custom slash commands

[rarean]

The idea is to borrow the custom slash commands concept from Claude Code and incorporate this functionality with AmzQ Dev.
Here is a repo with some custom claude code commands that provide more use-cases people have come up with: https://github.com/qdhenry/Claude-Command-Suite/tree/main/.claude/commands

I am not asking or implying that all of these custom commands are needed or wanted; rather, at a high level, the ability to create custom commands like these is the desired outcome.

While it is possible to add linting and standards for projects in AmzQ via "project rules", the rules don't allow you to chain different tools together to create a specific workflow with your specific tools for a given project. i.e. lint with black, ruff, then use isort and fix any errors in a Python project.

You can add this info to project rules, but the order in which the tools run and additional instructions to fix tests are not possible with the current /context command. Please correct me if I am wrong.

There are at least two other feature requests that I found related to this that may be solved with custom slash commands

• Feature Request: Amazon Q Personas for Role-Based AI Assistance #2131
• Feature request: Add /prompts create subcommand to create prompt from within CLI #1342

[rarean]

I added an MVP/POC version of what this would look like in a feature branch on my fork
main...rarean:amazon-q-developer-cli:feat-1342
The clippy.md example I added was created after CI tests failed from my first push with code changes. Thus, a real-world example of how custom slash commands could be used in the project.

After building and compiling locally, you can test the POC with:

• source "$HOME/.cargo/env"
• cargo run --bin chat_cli -- settings chat.enableCommands true
• cargo run --bin chat_cli -- chat

There are some example custom commands in the .amazonq/commands dir to test with. These can be run with:

• /project:code-review
• /project:test-command
• /project:clippy

Yes, the same thing could be accomplished with a make file or pre-commit, but those approaches are limited to simply running the command and not adding additional instructions to fix all the errors and reference design patterns or project rules when fixing the errors.