/my-command

Review the changes in this PR. Focus on:

1. **Correctness**: Does the logic do what it claims? 2. **Security**: Any injection risks, leaked secrets, or auth bypasses? 3. **Tests**: Are edge cases covered? 4. **Style**: Does it follow our conventions in CLAUDE.md?

For each issue found, classify as:

• **Must fix**: Bugs, security issues
• **Should fix**: Missing tests, unclear naming
• **Nit**: Style preferences, optional improvements

Start by running `git diff main...HEAD` to see all changes.

$ARGUMENTS

---

This is a starter command file. The filename `my-command.md` becomes the slash command `/my-command`.

Everything above the `---` is the prompt that Claude receives when you invoke the command. `$ARGUMENTS` is replaced with whatever you type after the command name.

For example, `/my-command focus on auth` would replace `$ARGUMENTS` with "focus on auth".

What Makes a Good Command

• A clear, specific task (not open-ended)
• Context about your project's standards or conventions
• Structured output format so results are consistent
• `$ARGUMENTS` to allow per-invocation customisation

Limitations

Commands are simple prompt templates. They do not support:

• Frontmatter (model overrides, invocation control, tool restrictions)
• Supporting files (scripts, references, assets)
• Dynamic context injection (`!` backtick syntax)
• Auto-loading based on task relevance

For any of those features, use a [skill](^Skills are the newer, more powerful replacement for commands. See the skills/ directory for the full structure) instead. Commands and skills create the same `/slash-command` interface.