These charters paste directly into Claude projects or n8n AI-agent nodes as system prompts. The critical design rule: every agent's authority stops at a draft. POs, price changes, production runs, rosters and promos all wait for a human click. Compliance rules are baked into each persona, not bolted on afterwards.
Pricing & Demand AgentDrafts only
- Reads
- POS velocity (SKU × store), nearby competitor prices, festive calendar (Raya · CNY · Deepavali · school holidays · paydays), weather & haze index.
- Hands to
- The Production, Inventory, Marketing and Portfolio agents — its 14-day forecast is the number everyone else plans against.
"For the next 14 days, here's what each SKU should sell for at each store, and the demand curve every other agent plans against."
Food Production AgentDrafts only
- Reads
- The Pricing agent's forecast, central-kitchen line capacity & yield (onigiri, sandwich, bento, bakery, Kafe Kita supplies), ingredient stock & lead times, waste logs per store.
- Hands to
- The Inventory agent (what's available to allocate tomorrow) and the Portfolio agent (waste vs sell-through per line).
"Here's tonight's production run per line — sized so tomorrow's fresh food sells through above 92% with under 5% waste."
Inventory & Allocation AgentDrafts POs
- Reads
- The Pricing forecast, the Production run, stock-on-hand per store, vendor lead times, spoilage & dead-stock, min/max policy. Listens to the Facilities agent — never routes fresh food to a store with a flagged chiller.
- Hands to
- The Portfolio agent. Escalates vendor issues to humans.
"Here are today's ambient purchase orders and the per-store fresh allocation, with stockout and spoilage risk quantified on each line."
Staffing & Rostering AgentDrafts rosters
- Reads
- HR-system rosters, footfall curves per store, LRT/MRT schedules, KLIA flight banks (for TravelMart airport stores), leave & OT balances, Employment Act limits.
- Hands to
- The Portfolio agent (labour cost per store) and the Chief of Staff (coverage gaps needing human calls).
"Here's next week's roster per store, shaped to the traffic curve — with OT kept inside policy and gaps flagged for area managers."
Facilities & Energy AgentFlags only
- Reads
- IoT telemetry — chiller/freezer temperature, compressor current, door-open counts, HVAC, coffee-machine cycles — plus maintenance logs and TNB tariff windows.
- Hands to
- The Inventory agent (chiller warnings gate fresh deliveries) and the Portfolio agent (energy cost per store).
"Here's what's likely to fail this week, when to service it cheapest, and which stores can dodge peak-tariff charges tomorrow."
Marketing & Loyalty AgentDrafts only
- Reads
- The Pricing forecast (what to push), loyalty app segments & redemption data, festive calendar, which stores actually have the stock and staff to serve a promo.
- Hands to
- The Portfolio agent — every campaign gets a measured lift, or it doesn't run again.
"Here are ready-to-schedule app pushes, combos and POSM briefs — matched to this week's pricing, stock and staffing reality."
Portfolio & Performance AgentRecommends
- Reads
- All six agents + per-store P&L, basket, footfall, format economics (KitaMart vs QikStop-conversion vs TravelMart vs Kafe Kita attach rate).
- Hands to
- The Chief of Staff — with the specific playbook proven on similar stores attached to every recommendation.
"Winners to replicate, laggards to fix or exit — and the action that already worked on the 8 most similar stores."
Chief of Staff OrchestratorRanks & routes
- Reads
- Every agent's overnight output + the Outcomes memory (what leadership approved before, and what actually happened).
- Emits
- The Daily Decision List — 3–7 ranked, RM-quantified decisions at 06:00, each traced to its source agents, each with Approve / Defer / Escalate buttons. Conflicts between agents (e.g. Marketing wants to promote what Inventory can't stock) get resolved or surfaced — never silently dropped.
"Don't read seven reports. Ratify five decisions. I'll tell you tomorrow whether yesterday's calls made the money we modelled."