Supply chain core
Hub-and-spoke agent-skill clusters, one per stack (Astro·GSAP·Remotion, Tauri, …). Installable via skills.sh.
npx -y skills add Sheshiyer/skill-clusters --skill supply-chain-coreAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Shared reference for the supply-chain cluster: the service-level vs landed-cost trade-off that every function optimizes, the SCOR stage map, the buffer/constraint model, common units (OTIF, fill rate, lead time, total landed cost), and the escalation matrix. USE WHEN setting safety stock, accepting a freight rate, sequencing production, or making a disposition/clearance call — the one objective all spokes share.
SKILL.md
6.5 KB, as published. Nobody here has run it
Supply-Chain Core
Shared model for the supply-chain cluster. Planning, make, deliver, and return spokes all
optimize the same objective with different levers — keep the trade-off, the units, and the
escalation rules consistent here so no spoke contradicts another.
1. The decision this cluster turns on
Every supply-chain choice is the same trade-off: protect the customer promise at the lowest total landed cost, given a binding constraint.
maximize Service (OTIF / fill rate / on-time)
minimize Total landed cost (unit + freight + duty + inventory-holding + risk)
subject to the binding constraint of the stage (capacity · bottleneck · lead time · duty · demand variability)
buffered by a deliberate reserve (safety stock · capacity slack · lead-time pad · QA hold)
- Service is the promise you must not silently break — measured as OTIF, fill rate, on-time delivery, or quality acceptance.
- Total landed cost is the whole cost, not unit price: freight, duty, holding, expedite, scrap, and the cost of a stockout. Cheaper-per-unit that blows up landed cost or service is a false saving.
- The binding constraint differs by stage — name it before you optimize, because you can only buy service by spending on the active constraint.
- The buffer is how you absorb variability. Every buffer (safety stock, capacity reserve, lead-time pad, inspection hold) is a priced decision: more buffer = more service + more cost. State it.
Rule: never move service for unit-cost (or unit-cost for service) without saying so; always identify the binding constraint first; treat any buffer change as a cost-and-risk decision worth naming.
2. The SCOR backbone (which spoke owns the stage)
PLAN ───────> SOURCE ──────> MAKE ──────────> DELIVER ─────────> RETURN
│ │ │ │ │
inventory- energy- production- carrier-relationship- returns-
demand- procurement scheduling + management + reverse-
planning (utilities) quality- logistics-exception- logistics
nonconformance management +
customs-trade-compliance
- Plan sets the demand signal and buffers that everything downstream consumes.
- Source secures inputs (materials via the buyer; utilities via
energy-procurement). - Make converts inputs (sequence + bottleneck via scheduling; conformance via quality).
- Deliver moves goods (capacity + rates; in-transit recovery; cross-border clearance).
- Return brings goods back and recovers value (disposition, refund, warranty).
3. Shared conventions & units
- Service metrics: OTIF (on-time-in-full) is the default promise; fill rate for inventory; on-time % for transit; first-pass yield / acceptance for quality. Quote the metric, not a vibe.
- Cost: always total landed cost (unit + freight + duty/tariff + holding + risk/expedite). Inventory holding ≈ 18–28%/yr of value unless told otherwise.
- Time: lead time = order → receipt; cycle time = start → finish of one unit/job; quote mean and variability — buffers exist to cover the variability, not the mean.
- Buffers: safety stock (inventory), capacity reserve (make), lead-time pad (deliver), QA/inspection hold (quality). One word for the same idea: deliberate slack against variability.
- Hard stops (never trade away): restricted-party / sanctions hits, safety or regulatory quality holds, missing legally-required documentation. These are gates, not optimizations.
4. Cross-cutting matrix (stage × constraint × buffer × spoke)
| Stage | Typical binding constraint | The buffer you tune | Service metric | Spoke |
|---|---|---|---|---|
| Plan | Demand variability | Safety stock | Fill rate | inventory-demand-planning |
| Source (utilities) | Price/tariff exposure | Hedge / contract term | $/unit stability | energy-procurement |
| Make — schedule | Bottleneck capacity | Capacity reserve, WIP | Schedule attainment | production-scheduling |
| Make — quality | Defect/escape rate | Inspection / QA hold | First-pass yield | quality-nonconformance |
| Deliver — capacity | Carrier capacity | Backup carriers, slack | Tender acceptance, OTIF | carrier-relationship-management |
| Deliver — recovery | Exception/dwell time | Lead-time pad, expedite | On-time recovery | logistics-exception-management |
| Deliver — cross-border | Duty / clearance time | Broker, FTA, bonded buffer | Clearance on-time | customs-trade-compliance |
| Return | Recovery value vs cost | Disposition policy | Recovery rate, cycle | returns-reverse-logistics |
5. Escalation discipline (shared)
Each spoke defines its own thresholds; the shared rule is escalate before you absorb risk quietly. Trip an escalation when: a buffer breaches its floor (stockout imminent, no backup carrier, capacity overcommitted), a hard stop fires (compliance/quality/safety), landed cost or service deviates beyond plan tolerance, or a single decision exceeds the function's authority (rate increase, write-off, disposition value, duty exposure). State the trade-off and the constraint when you escalate — the decision is the same shape regardless of stage.
6. Shared guardrails
- Protect the promise: don't trade OTIF/fill-rate for unit savings without naming it (and vice versa).
- Name the binding constraint before optimizing — you can only buy service on the active constraint.
- Price every buffer: safety stock, capacity reserve, lead-time pad, QA hold are cost-and-risk decisions, not free.
- Total landed cost, always — never let cheaper-per-unit hide freight, duty, holding, or stockout cost.
- Honor hard stops: restricted-party, safety/quality holds, regulatory docs are gates, not trade-offs.
- Escalate on threshold breach, don't absorb risk silently; carry the trade-off + constraint into the escalation.