cc-os/plugins/cc-architect/workflows/create-command.md

status	note	model_tier
legacy	For maintenance of existing commands only. Prefer skills for new functionality.	requirements design implementation main main haiku

Workflow: Create Command (LEGACY)

LEGACY WORKFLOW: This workflow is for maintaining existing slash commands. For new functionality, prefer creating skills via the create-skill workflow instead.

When to use

• Creating a new slash command (legacy maintenance only)
• Concept is clear and requirements understood

Before you start

If concept unclear, brainstorm: references/brainstorming/workflow.md

Pattern consideration

See references/tool-patterns/catalog.md.

Question	If Yes
Is command concept unclear?	Brainstorm first
Will command modify files?	Include verification in instructions
Will command wrap a skill?	Skill handles patterns; command just invokes

---

Inputs

Input	Example
command_name	create-backup (kebab-case)
command_purpose	"Create a backup of the current project"
target_plugin_path	/home/user/my-plugin

Subtasks

1. Understand requirements

Subagent: Main with user.

Do:

• What should this command do?
• What tools needed? (Read, Write, Bash, etc.)
• Wrap existing skill or standalone?
• What inputs from user?

2. Design the command

Subagent: Main.

Do:

• Determine command name (kebab-case, descriptive)
• Write one-line description
• List required tools
• Outline instructions
• If modifies files: include verification step

A command file has two parts: YAML frontmatter (description, optional allowed-tools) and a body of plain instructions the model follows when the command is invoked. See ${CLAUDE_PLUGIN_ROOT}/commands/cc-architect.md for a minimal real example, and ${CLAUDE_PLUGIN_ROOT}/templates/config-command.tmpl for one with input parsing.

3. Write the command file

Subagent: haiku. Read: ${CLAUDE_PLUGIN_ROOT}/references/layout/descriptions.md

Do:

1. Generate the description using the generate-description workflow (${CLAUDE_PLUGIN_ROOT}/workflows/generate-description.md)
   • Input: type=command, name, purpose, triggers
2. Create: {target_plugin_path}/commands/{command_name}.md
3. Write:
   • Frontmatter (description, allowed-tools if the command needs specific tools)
   • Step-by-step instructions in the body
   • Any user-request parsing guidance the command needs

Validation

• Filename matches command name
• Description follows formula: "[Action verb] [what] when [trigger]"
• Only necessary tools listed
• Instructions clear and actionable
• Command does one thing well

Output

{target_plugin_path}/commands/{command_name}.md