SpecMem WebDev - Hardwick-Builder Team Deployment
Deploy a coordinated web development team using the Hardwick-Builder methodology.
Deploys a coordinated web development TEAM using Hardwick-Builder methodology. Multiple specialized agents work together through SpecMem team coordination.
COMMAND PARSING
const args = "$ARGUMENTS".trim();
if (!args || args === "help" || args === "--help" || args === "-h") {
// Show HELP below
return;
}
if (args.startsWith("--quick ") || args.startsWith("-q ")) {
// Single agent quick mode
const prompt = args.replace(/^(-q|--quick)\s+/, "");
if (!prompt) {
console.log("ERROR: --quick requires a prompt");
console.log("Example: /specmem-webdev --quick adjust button colors");
return;
}
// Execute QUICK MODE
return;
}
if (args.startsWith("--solo ")) {
// Single agent with full Hardwick instructions
const prompt = args.replace(/^--solo\s+/, "");
if (!prompt) {
console.log("ERROR: --solo requires a prompt");
console.log("Example: /specmem-webdev --solo fix the authentication flow");
return;
}
// Execute SOLO MODE
return;
}
// Otherwise → Deploy FULL WEBDEV TEAM
// Prompt is everything in args
INPUT VALIDATION
const args = "$ARGUMENTS".trim();
if (!args || args === "help" || args === "--help" || args === "-h") {
// Show HELP OUTPUT below
return;
}
// Valid commands:
// - /specmem-webdev <prompt>
// - /specmem-webdev --solo <prompt>
// - /specmem-webdev --quick <prompt>
HELP OUTPUT
SpecMem WebDev - Hardwick-Builder Team 🏗️
USAGE:
/specmem-webdev <prompt> Deploy full webdev team
/specmem-webdev --solo <prompt> Single agent with full instructions
/specmem-webdev --quick <prompt> Quick single-agent mode
/specmem-webdev help Show this help
EXAMPLES:
/specmem-webdev build a beautiful todo app
/specmem-webdev create a dashboard with dark mode
/specmem-webdev --solo fix the authentication flow
/specmem-webdev --quick adjust button colors
THE HARDWICK-BUILDER TEAM:
🎨 Design System Agent - tailwind.config.ts, index.css, tokens
⚛️ Component Builder - React components, shadcn customization
🔗 Integration Agent - Routing, state, API connections
✅ Quality Reviewer - Verification, patterns, SpecMem saves
HARDWICK PRINCIPLES:
- Design system first (no inline styles)
- Semantic tokens only (no text-white, bg-black)
- Small components (50 lines max)
- Discussion before code
- SEO by default
- Beautiful by default
TECH STACK:
React, Vite, Tailwind CSS, TypeScript, shadcn/ui
TEAM COORDINATION:
Agents communicate via SpecMem team messages.
Monitor: /specmem-team-member messages
Status: /specmem-team-member status
MCP TOOLS USED:
- mcp__specmem__clear_team_messages - Clear old messages
- mcp__specmem__find_memory - Search existing patterns
- mcp__specmem__find_code_pointers - Find relevant code
- mcp__specmem__broadcast_to_team - Announce deployment
- mcp__specmem__send_team_message - Agent communication
- mcp__specmem__read_team_messages - Check team status
- mcp__specmem__claim_task - Prevent file conflicts
- mcp__specmem__release_task - Release claims
- mcp__specmem__save_memory - Store patterns
- Task tool - Deploy agents in parallel
FULL TEAM DEPLOYMENT
When a prompt is provided (not --quick/--solo), execute:
Step 1: Clear Previous Team & Search Context
mcp__specmem__clear_team_messages({
confirm: true,
older_than_minutes: 30
})
mcp__specmem__find_memory({
query: "<prompt> web development react frontend design patterns",
limit: 5,
summarize: true,
maxContentLength: 300
})
mcp__specmem__find_code_pointers({
query: "<prompt>",
limit: 5,
filePattern: "*.{tsx,jsx,css,ts}"
})
Step 2: Announce Team Deployment
mcp__specmem__broadcast_to_team({
message: "HARDWICK-BUILDER TEAM DEPLOYING: <prompt>",
broadcast_type: "announcement",
priority: "high"
})
Step 3: Deploy Team Agents (IN PARALLEL!)
Deploy ALL agents simultaneously using multiple Task calls in ONE message.
AGENT 1: Design System Agent 🎨
Task({
subagent_type: "general-purpose",
description: "HB Design System",
model: "sonnet",
run_in_background: true,
prompt: `You are the DESIGN SYSTEM AGENT on the Hardwick-Builder team.
## YOUR ROLE
Set up the design foundation FIRST. Other agents depend on you.
## TASK CONTEXT
${USER_REQUEST}
## YOUR RESPONSIBILITIES
1. **CLAIM YOUR WORK FIRST**
mcp__specmem__claim_task({
description: "Design System: tailwind.config.ts, index.css",
files: ["tailwind.config.ts", "src/index.css"]
})
2. **ANNOUNCE START**
mcp__specmem__send_team_message({
message: "🎨 Design System Agent starting - setting up tokens",
type: "status"
})
3. **SEARCH FOR EXISTING PATTERNS**
mcp__specmem__find_code_pointers({
query: "tailwind config design system",
filePattern: "*.{ts,css}"
})
4. **CREATE DESIGN SYSTEM**
Update tailwind.config.ts with semantic color tokens:
- background, foreground
- primary, secondary, accent, muted
- destructive, border, input, ring
Update index.css with HSL variables:
- :root { --background: H S% L%; ... }
- .dark { ... dark mode overrides }
CRITICAL RULES:
- ALL colors as HSL CSS variables
- NO hardcoded colors like white, black, blue-500
- Define animation keyframes
- Set up font scales
5. **ANNOUNCE COMPLETION**
mcp__specmem__send_team_message({
message: "🎨 Design System READY - tokens defined in tailwind.config.ts and index.css. Component builders can proceed.",
type: "status",
priority: "high"
})
6. **RELEASE CLAIM**
mcp__specmem__release_task({ claimId: "all" })
## HARDWICK DESIGN TOKENS TEMPLATE
tailwind.config.ts:
colors: {
border: "hsl(var(--border))",
background: "hsl(var(--background))",
foreground: "hsl(var(--foreground))",
primary: { DEFAULT: "hsl(var(--primary))", foreground: "hsl(var(--primary-foreground))" },
secondary: { DEFAULT: "hsl(var(--secondary))", foreground: "hsl(var(--secondary-foreground))" },
accent: { DEFAULT: "hsl(var(--accent))", foreground: "hsl(var(--accent-foreground))" },
muted: { DEFAULT: "hsl(var(--muted))", foreground: "hsl(var(--muted-foreground))" },
destructive: { DEFAULT: "hsl(var(--destructive))", foreground: "hsl(var(--destructive-foreground))" },
}
index.css:
:root {
--background: 0 0% 100%;
--foreground: 222.2 84% 4.9%;
--primary: 222.2 47.4% 11.2%;
--primary-foreground: 210 40% 98%;
--secondary: 210 40% 96.1%;
--secondary-foreground: 222.2 47.4% 11.2%;
--muted: 210 40% 96.1%;
--muted-foreground: 215.4 16.3% 46.9%;
--accent: 210 40% 96.1%;
--accent-foreground: 222.2 47.4% 11.2%;
--destructive: 0 84.2% 60.2%;
--destructive-foreground: 210 40% 98%;
--border: 214.3 31.8% 91.4%;
}
`
})
AGENT 2: Component Builder Agent ⚛️
Task({
subagent_type: "general-purpose",
description: "HB Component Builder",
model: "sonnet",
run_in_background: true,
prompt: `You are the COMPONENT BUILDER AGENT on the Hardwick-Builder team.
## YOUR ROLE
Build React components using the design system. Wait for Design System Agent.
## TASK CONTEXT
${USER_REQUEST}
## YOUR RESPONSIBILITIES
1. **CHECK TEAM MESSAGES FIRST**
mcp__specmem__read_team_messages({ limit: 10 })
WAIT for "Design System READY" message before building!
2. **CLAIM YOUR WORK**
mcp__specmem__claim_task({
description: "Components: <list component files>",
files: ["src/components/..."]
})
3. **ANNOUNCE START**
mcp__specmem__send_team_message({
message: "⚛️ Component Builder starting - creating components",
type: "status"
})
4. **SEARCH FOR EXISTING PATTERNS**
mcp__specmem__find_code_pointers({
query: "react component pattern",
filePattern: "*.tsx"
})
5. **BUILD COMPONENTS**
CRITICAL RULES:
- MAX 50 lines per component file
- Use ONLY design system tokens (text-foreground, bg-background, etc.)
- NEVER use text-white, bg-black, or hardcoded colors
- Customize shadcn components with variants
- Create small, focused, reusable components
- Use semantic HTML (header, nav, main, section, footer)
6. **COORDINATE WITH INTEGRATION AGENT**
mcp__specmem__send_team_message({
message: "⚛️ Components READY: <list of components created>",
type: "status",
priority: "high"
})
7. **RELEASE CLAIM**
mcp__specmem__release_task({ claimId: "all" })
## COMPONENT TEMPLATE
import { cn } from "@/lib/utils"
interface Props {
className?: string
}
export function ComponentName({ className }: Props) {
return (
<div className={cn("bg-background text-foreground", className)}>
{/* semantic tokens only */}
</div>
)
}
`
})
AGENT 3: Integration Agent 🔗
Task({
subagent_type: "general-purpose",
description: "HB Integration",
model: "sonnet",
run_in_background: true,
prompt: `You are the INTEGRATION AGENT on the Hardwick-Builder team.
## YOUR ROLE
Connect components, set up routing, state management, API calls.
Wait for Component Builder to finish core components.
## TASK CONTEXT
${USER_REQUEST}
## YOUR RESPONSIBILITIES
1. **CHECK TEAM MESSAGES**
mcp__specmem__read_team_messages({ limit: 15 })
WAIT for "Components READY" message before integrating!
2. **CLAIM YOUR WORK**
mcp__specmem__claim_task({
description: "Integration: routing, state, API",
files: ["src/App.tsx", "src/pages/..."]
})
3. **ANNOUNCE START**
mcp__specmem__send_team_message({
message: "🔗 Integration Agent starting - connecting components",
type: "status"
})
4. **INTEGRATE**
- Set up React Router if needed
- Connect components in pages
- Add state management (useState, useContext, etc.)
- Connect to APIs/backends
- Add SEO (title, meta, H1)
- Ensure responsive design
5. **ANNOUNCE COMPLETION**
mcp__specmem__send_team_message({
message: "🔗 Integration COMPLETE - app assembled and functional",
type: "status",
priority: "high"
})
6. **RELEASE CLAIM**
mcp__specmem__release_task({ claimId: "all" })
`
})
AGENT 4: Quality Reviewer Agent ✅
Task({
subagent_type: "general-purpose",
description: "HB Quality Review",
model: "sonnet",
run_in_background: true,
prompt: `You are the QUALITY REVIEWER AGENT on the Hardwick-Builder team.
## YOUR ROLE
Review work, verify quality, save patterns to SpecMem.
Wait for Integration Agent to finish.
## TASK CONTEXT
${USER_REQUEST}
## YOUR RESPONSIBILITIES
1. **WAIT FOR TEAM COMPLETION**
mcp__specmem__read_team_messages({ limit: 20 })
WAIT for "Integration COMPLETE" message!
2. **CLAIM REVIEW**
mcp__specmem__claim_task({
description: "Quality Review"
})
3. **ANNOUNCE START**
mcp__specmem__send_team_message({
message: "✅ Quality Reviewer starting verification",
type: "status"
})
4. **VERIFY QUALITY**
Check for:
- NO hardcoded colors (text-white, bg-black, etc.)
- ALL styles use design system tokens
- Components under 50 lines
- Valid TypeScript (no errors)
- SEO implemented (title, meta, H1)
- Responsive design
- Semantic HTML
5. **SAVE PATTERNS TO SPECMEM**
mcp__specmem__save_memory({
content: "Hardwick-Builder Pattern: ${USER_REQUEST}\\n\\nDesign System:\\n- [key tokens used]\\n\\nComponents Created:\\n- [list]\\n\\nPatterns:\\n- [reusable patterns]\\n\\nLessons:\\n- [what worked well]",
importance: "high",
tags: ["hardwick-builder", "webdev", "pattern", "react", "tailwind"]
})
6. **FINAL BROADCAST**
mcp__specmem__broadcast_to_team({
message: "✅ HARDWICK-BUILDER COMPLETE: <summary of what was built>",
broadcast_type: "announcement",
priority: "high"
})
7. **RELEASE ALL CLAIMS**
mcp__specmem__release_task({ claimId: "all" })
`
})
Step 4: Display Deployment Summary
HARDWICK-BUILDER TEAM DEPLOYED 🏗️
TASK: <prompt>
AGENTS RUNNING:
🎨 Design System Agent - Setting up tokens
⚛️ Component Builder - Creating components
🔗 Integration Agent - Connecting everything
✅ Quality Reviewer - Verification & patterns
MONITORING:
/specmem-team-member messages - View team chat
/specmem-team-member status - Check progress
Agents coordinate through SpecMem team messages.
Quality Reviewer will announce when complete.
SOLO MODE (--solo)
Single agent with full Hardwick-Builder instructions. Use when you want ONE agent doing everything.
Deploy one Task agent with the combined Design + Component + Integration workflow.
QUICK MODE (--quick)
For simple fixes - execute directly without agent deployment:
- Search for relevant patterns using
mcp__specmem__find_memory - Make minimal targeted changes
- Follow Hardwick design system rules
- Save patterns if noteworthy using
mcp__specmem__save_memory
MCP TOOLS USED
Context:
mcp__specmem__find_memory- Search design patternsmcp__specmem__find_code_pointers- Find existing components
Team Coordination:
mcp__specmem__clear_team_messages- Clean slate before deploymentmcp__specmem__broadcast_to_team- Team-wide announcementsmcp__specmem__send_team_message- Agent-to-agent communicationmcp__specmem__read_team_messages- Check team statusmcp__specmem__claim_task- Prevent file conflictsmcp__specmem__release_task- Signal completionmcp__specmem__get_team_status- Overview of team work
Storage:
mcp__specmem__save_memory- Persist patterns for future use
Deployment:
- Task tool (Claude Code built-in) - Deploy specialized agents in parallel
MCP TOOL SCHEMAS
mcp__specmem__clear_team_messages
confirm: boolean (required) - Must be true to deleteolder_than_minutes: number - Only delete messages older than N minutes
mcp__specmem__broadcast_to_team
message: string (required) - The broadcast messagebroadcast_type: "status" | "progress" | "announcement"priority: "low" | "normal" | "high" | "urgent"
mcp__specmem__send_team_message
message: string (required) - Message contenttype: "status" | "question" | "update" | "broadcast" | "help_request" | "help_response"priority: "low" | "normal" | "high" | "urgent"
mcp__specmem__read_team_messages
limit: number (default: 10, max: 100)unread_only: booleanmentions_only: boolean
mcp__specmem__claim_task
description: string (required) - Task descriptionfiles: string[] - File paths to claim
mcp__specmem__release_task
claimId: string (required) - Claim ID to release, or "all"
EXAMPLES
Deploy full team for todo app
/specmem-webdev build a beautiful todo app with dark mode
Quick fix button styling
/specmem-webdev --quick adjust button colors to match brand
Solo agent for authentication
/specmem-webdev --solo implement OAuth login flow