Create Command: Scaffold New Slash Commands
Interactive agent that scaffolds new slash commands with best practices, proper frontmatter, and structured templates
- 💡 Use Case:
- Creating new slash commands for Claude Code with consistent structure and patterns
- 📤 Expected Output:
- A complete slash command file with frontmatter, context blocks, and instructions
Prompt Content
Create Command: Scaffold New Slash Commands
An interactive agent that guides you through creating well-structured slash commands for Claude Code. It asks the right questions and generates commands following proven patterns.
Purpose
Creating effective slash commands requires understanding frontmatter options, context gathering patterns, and instruction structures. This agent encapsulates best practices so you can create powerful commands quickly without memorizing the format.
Instructions
Step 1: Gather Requirements
Ask the user these questions:
Question 1: Command Category What category best describes this command?
| Category | Description | Defaults |
|---|---|---|
git-workflow | Git operations (commit, branch, push, merge) | model: haiku, git tools |
github | GitHub operations (PR, issues, CI) | model: haiku, gh + git tools |
quality | Linting, testing, type checking | test/lint tools |
debugging | Fix errors, investigate issues | full tools, decision trees |
code-transform | Refactoring, code generation | full tools, file references |
utility | General purpose automation | varies |
Question 2: Arguments Does this command need arguments?
| Option | Syntax | Example |
|---|---|---|
| None | (no arguments) | /lint |
| Single required | $1 | /review-pr 123 |
| Multiple positional | $1, $2, $3 | /deploy staging v1.2.3 |
| Free-form text | $ARGUMENTS | /branch add dark mode toggle |
Question 3: File References Does the command operate on specific files?
- Yes - Use
@$1syntax to reference file contents - No - Command works on general context
Question 4: Complexity How complex is the workflow?
| Level | Characteristics | Structure |
|---|---|---|
simple | Run command, report result | Linear steps |
linear | Sequential steps, no branching | Numbered instructions |
branching | Pre-checks, multiple paths | Decision tree |
complex | Multi-phase, verification, fallback | Full protocol with red flags |
Step 2: Generate Command File
Based on the answers, generate the command file with:
Frontmatter:
---
description: <imperative verb phrase>
argument-hint: <show expected args>
allowed-tools: <category defaults>
model: <haiku for routine, omit for complex>
---
Context Blocks using semantic XML tags:
<current_branch>
!`git branch --show-current`
</current_branch>
Structure Templates based on complexity level.
Step 3: Write the File
- Write the file to
.claude/commands/<name>.md - Show the user the generated command
- Explain how to use it
Expected Output
- Format: Complete markdown file with YAML frontmatter
- Key sections:
- Frontmatter with description, tools, model
- Context blocks with shell commands
- Decision tree (if branching/complex)
- Numbered instructions
- Red flags section (if complex)
Example
Input:
/create-command deploy "Deploy to staging or production environment"
Output:
A complete command file at .claude/commands/deploy.md with:
- Proper frontmatter for deployment category
- Context blocks for git status, current branch, latest tag
- Decision tree handling uncommitted changes
- Instructions for build and deploy steps
Done When
- User requirements gathered via questions
- Command file generated with correct frontmatter
- Context blocks use appropriate shell commands
- Instructions match complexity level
- File written to
.claude/commands/