--- name: create-command description: Create a new Brokkr WhatsApp command with proper structure disable-model-invocation: true user-invocable: true allowed-tools: Read, Write, Bash, Glob argument-hint: --- # Create Brokkr Command ## Overview Create a new command for the Brokkr WhatsApp bot following the established patterns. ## Arguments - `$0` - Command name (lowercase, hyphens allowed) - `$1` - (Optional) Command type: "claude", "skill", or "internal" ## Process ### Step 1: Validate Command Name Check the name: - Must be lowercase - Only letters, numbers, hyphens - Must start with a letter - Must not conflict with existing commands ```bash # Check for conflicts ls .brokkr\/commands\/ 2>\/dev\/null | grep -w "$0" && echo "CONFLICT" || echo "OK" ``` ### Step 2: Gather Requirements Ask the user: 1. **Description** - What does this command do? (for \/help, under 60 chars) 2. **Aliases** - Any shortcuts? (e.g., "r" for "research") 3. **Arguments** - What inputs does it need? - Required arguments - Optional arguments - Hint format (e.g., " [depth]") 4. **Handler type** - How does it work? - `claude` - Sends prompt to Claude (default category: `tasks`) - `skill` - Invokes a Claude Code skill (default category: `skills`) - `internal` - Calls built-in function (default category: `sessions`) 5. **Category** - (Optional) Override for `\/help` grouping: - `tasks` - Task commands - `sessions` - Session management - `scheduling` - Scheduled tasks - `skills` - Skill commands - `help` - Help commands ### Step 3: Create Command Directory ```bash mkdir -p .brokkr\/commands\/$0 ``` ### Step 4: Generate command.json Based on handler type, create the appropriate definition: **For Claude type:** (default category: `tasks`) ```json { "name": "$0", "description": "", "aliases": [], "category": "tasks", "priority": "CRITICAL", "source": "both", "arguments": { "required": [], "optional": [], "hint": "" }, "handler": { "type": "claude", "prompt": "" }, "session": { "create": true, "codeLength": 2 } } ``` **For Skill type:** (default category: `skills`) ```json { "name": "$0", "description": "", "category": "skills", "handler": { "type": "skill", "skill": "" }, "session": { "create": true, "codeLength": 2 } } ``` **For Internal type:** (default category: `sessions`) ```json { "name": "$0", "description": "", "category": "sessions", "handler": { "type": "internal", "function": "" }, "session": { "create": false } } ``` **Note:** The `category` field is optional. If omitted, it defaults based on handler type. Use explicit category to override (e.g., `"category": "scheduling"` for a claude-type command that schedules tasks). ### Step 5: Validate Run validation: ```bash node --input-type=module -e " import { validateCommand } from '.\/lib\/command-schema.js'; import { readFileSync } from 'fs'; const def = JSON.parse(readFileSync('.\/.brokkr\/commands\/COMMAND_NAME\/command.json')); const result = validateCommand(def); if (!result.valid) { console.error('Validation failed:', result.errors); process.exit(1); } console.log('Validation passed!'); " ``` (Replace COMMAND_NAME with the actual command name) ### Step 6: Test Registration ```bash node --input-type=module -e " import { CommandRegistry } from '.\/lib\/command-registry.js'; const registry = new CommandRegistry(); registry.discover(); const cmd = registry.get('COMMAND_NAME'); if (cmd) { console.log('Command registered:', cmd.name); console.log('Description:', cmd.description); } else { console.error('Registration failed!'); process.exit(1); } " ``` (Replace COMMAND_NAME with the actual command name) ## Argument Substitution Reference Use these placeholders in prompts: - `$ARGUMENTS` - All arguments as string - `$0`, `$1`, `$2` - Individual arguments by position - `${1:-default}` - Argument with default value - `${SESSION_CODE}` - Current session code ## Example Session User: `\/create-command summarize` Claude: Creating command "summarize". Let me gather some information: 1. What should this command do? (description for \/help) User: Summarize a URL or text content Claude: 2. Any aliases\/shortcuts? (comma-separated, or "none") User: sum, tldr Claude: 3. What arguments does it take? - Required: (comma-separated) - Optional: (comma-separated) - Hint format: (e.g., " [length]") User: Required: content. Optional: length. Hint: [short|long] Claude: 4. Handler type: claude, skill, or internal? User: claude Claude: 5. What prompt should Claude receive? (use $0 for content, $1 for length) User: Summarize the following content. Length: ${1:-medium}. Content: $0 Claude: Creating command... [creates files, validates, tests] Done! Command `\/summarize` is ready. Try: `\/summarize https:\/\/example.com short` ## Validation Checklist Before completing: - [ ] Name follows pattern: lowercase, letters\/numbers\/hyphens, starts with letter - [ ] Description under 60 characters - [ ] Category is valid: tasks, sessions, scheduling, skills, or help - [ ] command.json is valid JSON - [ ] Schema validation passes - [ ] No conflicts with existing commands - [ ] Registration test passes - [ ] Command appears in correct `\/help` category