Three sections below. Pantry is your raw data (Shopify, Meta, TikTok, Pendo). Prep table is the weekly briefs Claude assembles for you. Plate is what lands in front of you and Sam Monday morning. Click any card for plain English.
The mental model in one picture. Your data moves left to right: raw sources feed weekly silver platters, platters feed the briefs and drafts Claude writes for you, and those land in the channels you actually use (Slack, your inbox, your iPad). The AI layer sits on top of the data prep, never the other way around. Hover any line to trace it. Click a node to open the details.
If a Pantry box has no line leaving it, that data is sitting unused, those are usually the loudest quick wins. Dashed boxes on the right are channels we suggest adding.
SlabHaus, these are the briefs and drafts your team can stop building manually. Each one shows the ingredients, the time saved, and the back-and-forth Claude does on your behalf.
Marco wakes up Monday and sees last week's net + margin without opening four spreadsheets.
Sam goes live with the right tier focus, current inventory in hand, and 2-3 customer quotes ready as social proof.
Refund emails get drafted replies within an hour, with tier-appropriate language. Owner approves and sends.
Marco gets a Tuesday brief saying 'last week's strongest customer quote was X, here's the ad creative seed.' No more guessing what to test.
Use the customer-voice signal + Shopify order history to send abandoned-cart nudges that match the customer's tier, instead of one-size-fits-all email blasts.
Daily morning ping that flags any campaign whose spend pacing is breaking the plan, before it costs Marco the weekend.
SlabHaus, here's the order to build this in. Each step builds on the last, total: About a focused weekend (8-10 hours total). Step 1 unblocks everything. Step 2 is the biggest investment but pays back week one.. If a step looks like a cliff, hand it to @claude-code-guide.
Install the Claude Code CLI. It's the tool that runs everything below. If you've never opened Terminal, that's fine, this is the only command-line moment until you're set up.
npm install -g @anthropic-ai/claude-code claude --version # should print a version number
claude --version in Terminal prints a version number instead of an error.
Setup time: 5 minutes.
Drop a SessionStart hook into .claude/settings.json that converts any PDF / DOCX / XLSX / EML in data/raw_dropzone/ into markdown in data/converted/. This is your dopamine-hit step, 5 minutes, instant payoff.
Removes the hard wall of 'Claude can't read this' for vendor PDFs, refund email chains, and the live-stream schedule. Without this, every step that follows fails on the first PDF.
cp -r skill/silver-platter/scripts/templates/.. ~/.claude/hooks/ brew install pandoc poppler pip install xlsx2csv # then add the SessionStart entry to .claude/settings.json (template in business_os_hooks.md)
Setup time: 5 minutes
Write four ingest skills (one per source) that aggregate raw data into weekly summary markdown files. Schedule them via cron for Monday 6am. You can build these solo (1-2 weekends) OR hand the spec to @claude-code-guide and have all four in a day. Most operators take the handoff.
The silver platters are the 80%. Once they exist, every agent below has clean data to read instead of 30,000-row CSVs.
# write skills under .claude/skills/: # ingest_shopify_orders # ingest_meta_ads (extends your existing meta-ads-cli) # ingest_tiktok_ads # customer_voice_distill # and a master /weekly_silver_platters slash command # add cron: 0 6 * * MON cd ~/slabhaus && claude /weekly_silver_platters
Setup time: Half a day per skill (1-2 weekends DIY) OR ~1 day with @claude-code-guide handoff
Define four subagents under .claude/agents/. EA Orchestrator is the chief-of-staff Marco talks to. CFO/CMO/Ops are specialists, each scoped to one silver platter.
Now you have specialists who answer domain questions without you having to pre-frame the context. The orchestrator hierarchy keeps Marco from having to know which bot to call.
# write under .claude/agents/: # ea-orchestrator.md (model: opus, tools: [Read, Glob, Grep, Task]) # cfo-bot.md (model: sonnet, scoped to finance_weekly.md) # cmo-bot.md (model: sonnet, scoped to marketing + voice) # ops-bot.md (model: sonnet, scoped to ops_weekly.md) # Hand to @claude-code-guide for the templates if needed.
Setup time: 1-2 hours total
This is the only step that protects you from the AI shipping something you didn't see. PostToolUse hook on Edit|Write appends every action to outputs/audit_log.md. Stop hook nudges Marco at session end if the latest weekly report is unsigned. Skip it and you're flying blind.
You have receipts for everything the AI did. Nothing ships to customers (refund emails, marketing reports) without your sign-off. This is the trust layer.
# Templates already in demos/01_ecommerce_mystery_boxes/SETUP_PATH/business_os_hooks.md. # Paste the three hook entries into .claude/settings.json # and chmod +x the three hook scripts in .claude/hooks/.
Setup time: 10 minutes (hooks are already written)
Write three slash commands under .claude/commands/: /weekly_finance, /weekly_marketing, /refund_drama. Each triggers the right agent chain for a recurring Marco motion. Right now you ask the AI in full sentences every Monday. After this, you type /weekly_finance and walk to the kitchen.
Slash commands turn Marco's recurring weekly motions into one-keystroke triggers. He stops 'asking the AI' and starts running plays.
# write under .claude/commands/: # weekly_finance.md # weekly_marketing.md # refund_drama.md (one-off triage when refunds spike) # Pattern in demos/01_ecommerce_mystery_boxes/after/.claude/commands/.
Setup time: 30 minutes for all three
Next: hand @claude-code-guide a one-line ask for your next skill, or join the Early AI Dopters community to see how other operators evolved theirs.