cli: Skill for Claude Code

CLI Command Audit Skill

When to Use This Skill

Invoke this skill when the user:

• Asks to "review a command", "check a command", or "audit a command"
• Wants to verify a command conforms to project conventions
• Uses the slash command /review-command

---

Arguments

Arguments may be provided via $ARGUMENTS or gathered interactively.

Expected arguments:

• resource - Resource name (e.g., collection, study, workspace)
• command - Command name (e.g., list, get, create, update). Omit to review all commands under the resource.

If arguments are missing, use the ask_user tool to gather them.

---

Review Checklist

Work through each section below. For each item, check the relevant source files in cmd/{resource}/ and ui/{resource}/. Report findings at the end grouped by severity:

• 🔴 Must fix — non-conformant, will cause inconsistent UX or compile errors
• 🟡 Should fix — diverges from convention, not blocking
• 🟢 OK — conforms to the pattern

---

1. Output Flag (list commands only)

What to check: Every list command must register the output flags via shared.AddOutputFlags.

Expected pattern in cmd/{resource}/list.go (or equivalent list file):

// Options struct must embed OutputOptions
type ListOptions struct {
    Output shared.OutputOptions
    // ... other fields
}

// At the bottom of the constructor, after all other flags:
shared.AddOutputFlags(cmd, &opts.Output)

How to verify:

1. Read the command file
2. Confirm shared.OutputOptions is present in the options struct
3. Confirm shared.AddOutputFlags(cmd, &opts.Output) is called in the constructor
4. Confirm shared.ResolveFormat(opts.Output) drives the rendering switch (see §3)

Why it matters: Without this, users cannot use --json, --csv, --table, or -n flags, which breaks scripting and pipeline use cases.

---

2. Dependency Injection

What to check: Command constructors must use dependency injection — never construct a real client or writer internally.