Slash Command Creator

Create Claude Code slash commands following best practices for structure, frontmatter, and organization.

Shared conventions: See references/creator-conventions.md for the standard creator workflow (gather, generate, quality self-check, validate, review gate, commit).

Prompting best practices: See references/prompt-engineering-guide.md for optimization techniques to apply when generating command prompt content.

Workflow

1. Gather Requirements

Use AskUserQuestion for discrete choices, free-form for open-ended questions.

Command purpose (free-form):

• What task should this command automate?
• What problem does it solve?

Command location — AskUserQuestion with header "Location":

• "Project" (.claude/commands/ — versioned with the repo)
• "User Global" (~/.claude/commands/ — available across all projects)
• "Plugin" (commands/ — if adding to a prove-style plugin)

Arguments — AskUserQuestion with header "Arguments":

• "No arguments" (standalone command)
• "Single argument" (use $ARGUMENTS)
• "Multiple arguments" (use $1, $2, etc.)

Tool restrictions — AskUserQuestion with header "Tools":

• "Full access (Recommended)" (omit allowed-tools)
• "Read-only" (Read, Grep, Glob)
• "Git-only" (Bash(git *))
• "Custom" (let user specify)

Skill delegation (conditional — only if the command is part of a plugin):

• Which skill does it wrap? Plugin commands should be thin entry points that load the skill.

2. Generate the Command File

Refer to references/best-practices.md for frontmatter fields, tool restriction patterns, and argument handling.

Standard command structure:

---
description: [Action-oriented description — REQUIRED for model invocation]
argument-hint: [Expected arguments shown in autocomplete]
allowed-tools: [Tool restrictions if needed]
---

[Command prompt content here]

Plugin command structure (thin wrapper — keep all logic in the skill):

---
description: Do the thing
argument-hint: "[input]"
---

# Command Name: $ARGUMENTS

Load and follow the skill (`skills/<name>/SKILL.md` from the workflow plugin).

Apply the quality self-check from references/creator-conventions.md before presenting. Key checks for commands:

• Description front-loads the action verb (it's truncated at ~250 chars in autocomplete)
• Body doesn't restate the description (it's already in context)
• Plugin commands are thin — no business logic in the command file

3. Place the File

Location	Path	Notes
Project	.claude/commands/<name>.md	Versioned with repo
User	~/.claude/commands/<name>.md	Available across projects
Plugin	commands/<name>.md	Add scope to .claude/.prove.json
Namespaced	commands/<ns>/<name>.md	Invoked as /ns:name

4. Validate

Before writing the file, verify:

• Filename is lowercase, hyphen-delimited, matches the command's purpose
• description field is present (required for model invocation and /help)
• argument-hint is set if the command accepts arguments
• Tool restrictions match the command's intent (read-only for analysis, etc.)
• Plugin commands delegate to a skill — no business logic in the command file
• Instructions use numbered steps for multi-step workflows

Use AskUserQuestion with header "Review" to confirm: "Create Command" / "Revise".

Committing

Delegate to the commit skill. Example: feat(commands): add review command

Resources

• references/best-practices.md — frontmatter fields, tool restriction patterns, argument handling
• references/creator-conventions.md — shared creator workflow patterns
• references/prompt-engineering-guide.md — prompting techniques for command content