warehouse-shipment-planning

Files (3)

• SKILL.md
• .gitkeep · scripts/
• README.md · scripts/

SKILL.md (first lines)

# Warehouse shipment planning

This doc is responsible for:

- After each surgery day, computing a shipment plan for the upstream warehouse from the un-used remainder in the formal Tencent Docs inventory.
- Aligning "target stock band" and "current remainder" into one shared view for downstream notification and SPD shipping.
- Producing the unified baseline that `order-shipment` will consume.

When to use:

- User says "compute today's shipment plan", "what's left in stock", "how much more do we need to ship".
- User wants to notify the upstream warehouse based on the inventory's `blank-patient / blank-hospitalization-id` remainder.
- User wants the chain `warehouse → me → hospital SPD` to start with planning before SPD execution.

This doc is NOT responsible for:

- Clicking shipment / supply-confirm / upload-attachment in the SPD page.
- Direct edits to the Tencent Docs body.
- Pinning every range-bounded SKU to a fixed value when the warehouse-specific rule has not been stated.

Business chain in scope:

`upstream warehouse → me (notify warehouse, fill SPD shipment system, fill Tencent Docs inventory) → hospital SPD`.

Default input:

- Formal Tencent Docs inventory: `https://docs.qq.com/sheet/<prod-inventory-doc-id>`
- Body rows with `patient_id = blank` represent "un-used remainder stock".

Default output:

- A shipment-plan text aimed at the warehouse operator, with at minimum: `current_remainder`, `target_stock`, `suggested_replenishment`.

Shortest example: "from the inventory's blank-patient remainder, give me today's warehouse shipment plan."

## Default read semantics

- Default to reading the formal (prod) Tencent Docs inventory, not the test sheet.
- `patient_id = blank` is the operational definition of un-used remainder.
- Aggregate by `spec / model`; never mix in rows that already carry a patient or surgery date.
- `cath lab on-hand stock` must satisfy both `patient_id = blank` AND `storage_location = cath lab`. If the recent target-model batch is still tagged `company warehouse`, list it separately as "not counted in cath lab on-hand", never silently exclude.
- Through-browser `SpreadsheetApp` reads must do a one-shot stability re-check on target models. Any change between `cath lab / company warehouse` counts → take the latest read and note the variance in the result.
- Local evidence artifact must save full blank-patient rows with `row1 / spec / code / patient_id / storage / remark` — not only the `cath lab` slice — so later we can trace which rows were excluded by `storage_location`.
- Read-only inventories prefer the `tencent-docs-api-v3` shared backend; do not disturb the visible browser unless explicitly required.
- If reading via browser profile and the page shows `D0005 / please back up / table error / pending sync after network restore / offline`, do not trust the read even if `SpreadsheetApp` still returns a row model. Recover the page per `medical-inventory-tencent-import`'s recovery chain first, then re-read.
- For "what do we need to ship today" questions, also read the same-day schedule (surgery type + surgeon at minimum), not stock alone.

## Schedule coupling

- For "what do we still need today" plans, default to reading today's `<vendor-x>` case count and surgery types, plus surgeons, before deciding defaults.
- Schedule source priority:
  - Already-written Google Sheet cathlab records.
  - If today's records are missing / abnormal / the user wants the originals, fall back to schedule images / Word docs.

## Per-procedure default SKU bundles (excerpted, desensitized)

This skill encodes a per-procedure / per-surgeon default SKU bundle for cathlab ablation work. The full table covers ~9 procedure types × ~3 surgeons. A representative sample is shown below to give the shape; the full bundle lives in the private repo.

- **`<procedure-AF>`**:
  - Always include: `<sku-SF-A>`, `<sku-patch>`, `<sku-surface>`, `<sku-mapping-A>`, `<sku-mapping-B>`
  - `<sku-Deca>` depends on surgeon:
    - `<surgeon-A>`: include by default
    - `<surgeon-B>`: include by default
    - `<surgeon-C>`: exclude by default
- **`<procedure-AF-ultrasound>`**:
  - Same as `<procedure-AF>` plus `<sku-ice>`
- **`<procedure-PVC>`**:
  - If the surgery already finished and inventory has the actual deduction, defer to actuals; do not re-guess `<sku-Deca>`.
  - Reference bundle:
    - Always include: `<sku-SF-A>`, `<sku-patch>`, `<sku-surface>`
    - `<sku-Deca>` depends on surgeon:
      - `<surgeon-A>`: exclude by default (per recent 4-month actuals)
      - Other surgeons: include by default if left-side PVC is plausible pre-op
- **`<procedure-VT>`**: no auto default; agent must report `surgery_type / surgeon / current stock / suspected branches` and wait for human judgment.
- **`<procedure-SVT>`, `<procedure-WPW>`**:
  - Always include: `<sku-SF-A>`, `<sku-patch>`, `<sku-surface>`
  - `<surgeon-B>`: also include `<sku-Deca>`, `<sku-quad>`
- **`<procedure-EPS>`**: no auto default; agent reports and waits.
- **`<procedure-AT>`, `<procedure-AFL>`, `<procedure-PAC>`**:
  - Always include: `<sku-SF-A>`, `<sku-patch>`, `<sku-surface>`, `<sku-mapping-A>`
  - `<surgeon-A>` / `<surgeon-B>`: also include `<sku-Deca>`
- **Compound procedures** (e.g. `<procedure-AF>+<procedure-VT>`):
  - Union of component bundles. No "pick one" simplification. At minimum, the VT branch's `<sku-quad>` is also flagged.

So stock-deficit / replenishment decisions default to `surgery_type + surgeon`, not procedure name alone.
Any unknown surgeon → flag as `pending rule`, do not extrapolate from prior cases.

## Default report order

When the user asks for "current stock" or "by SKU":

- Report by `spec` / `model`, not commodity name first.
- Fixed order: `<sku-SF-A> → <sku-SF-B> → <sku-Deca> → <sku-patch> → <sku-surface> → <sku-mapping-A> → <sku-mapping-B> → <sku-quad> → <sku-ice>`.
- Verbal aliases the user actually uses: `SF (SF-A → SF-B)`, `Deca`, `patch`, `surface`, `mapping-A`, `mapping-B`, `quad`, `ice`.
- `SF` must be reported split into `SF-A → SF-B`, not only the combined total.
- BUT `SF` deficit / replenishment math is on the combined `SF-A + SF-B` total minus today's projected big-ticket consumption — do not fragment `7 + 1` into two independent deficits.
- Do not append commodity name / storage_location unless the user asks.

## Confirmed target stock bands

| SKU | Target total | Default composition |
| --- | --- | --- |
| SF | 8 | `<sku-SF-A> × 7`, `<sku-SF-B> × 1` |
| Deca | 5–6 | `<sku-Deca> × 5–6` |
| patch | 7–8 | `<sku-patch> × 7–8` |
| surface | 5–6 | `<sku-surface> × 5–6` |
| mapping-A | 5–6 | `<sku-mapping-A> × 5–6` |
| mapping-B | 5–6 | `<sku-mapping-B> × 5–6` |
| quad | 3–5 | `<sku-quad> × 3–5` |
| ice | 2–3 | `<sku-ice> × 2–3` |

## Confirmed boxing rules

- If `SF` and `Deca` share dimensions: 3 layers per box; 3 SF/Deca-sized slots per layer; 9 slots per box.
- Goal is "fill the box" (reduce in-transit crushing), not "fewest boxes".
- Confirmed dimensional conversions:
  - `1 × <sku-mapping-A> = 2 × <sku-patch>`
  - `1 × <sku-patch> = 2 × <sku-quad>`
- Confirmed per-layer combos:

From gricn/clinical-ops-skill-kit, MIT.

More from gricn/clinical-ops-skill-kit