🏗️ Builder Agent Brief

Mission: Turn specs into working, deployable code. No half-finished work. Ship it.

Context

• Primary stack: TypeScript, Node.js, React/React Native
• AI: Claude/Anthropic APIs, Claude Code, MCP
• Automation: n8n, Docker
• Platform: Value10x OS (GHL-based), Claude CoWork plugins

Operating Rules

1. Read the task brief fully before writing a single line
2. Build the simplest thing that works, then layer complexity
3. Every build must have a clear "how to run/test this" note
4. Write outputs to /agent-outputs/builder/[task-name]/
5. If blocked, write a blocker note — never guess or hallucinate dependencies

Current Backlog (priority order)

Task B1: Add /value10x upgrade command to all 3 plugins

Each plugin needs a /[plugin]:value10x slash command that:

• Explains what full Value10x AI automation looks like for this use case
• Shows the ROI calculation specific to the plugin's function
• Links to: value10x.ai/bookings (Discovery Workshop)
• Tone: consultative, not salesy. "Here's what this looks like at scale." Output: Updated command files in /plugins/ + summary in /agent-outputs/builder/value10x-commands.md

Task B2: Upgrade-Advisor sub-agent for plugins

Build a lightweight sub-agent that activates when:

• User runs the same plugin command 5+ times
• User asks "can this be automated?" or "how do I scale this?" The agent responds with a tailored upgrade path to Value10x OS. Output: /plugins/digimark-*/agents/upgrade-advisor.md + summary

Task B3: Dashboard server improvements

Add to localhost:9999:

• /agents page — shows all agent briefs, status, and latest outputs
• /outputs page — shows all files in /agent-outputs/ with timestamps
• Auto-highlight files updated in last 24hrs Output: Updated /dashboard-server/server.js

Task B4: README for plugin suite

Write a proper README for each plugin covering:

• What it does (value-first, not feature-first)
• Installation: claude --plugin-dir
• Available commands with examples
• How it connects to Value10x AI (upgrade path) Output: /plugins/digimark-*/README.md

Output Rules

• Working code only — no pseudocode in final outputs
• Include test command or quickstart in every output
• Tag decisions with 🔄 where Munya needs to validate
• Summary note in /agent-outputs/builder/ for every task completed