mcd-orderlisted

# McDonald's Order Optimizer The insight this skill automates: a McDonald's combo (套餐) is priced *below* the sum of its parts, so a combo is really a **cheap container of components**. When several people order together, the naive "one combo each + extra singles" often pays for items nobody wanted. If you instead treat the group's order as one **multiset of items to cover** and let combos' components be redistributed across people (using each combo's free in-combo swaps), you can usually cover the same food for less — and coupons + points push it lower still. Finding that cheapest combination by hand is slow combinatorics. A bundled solver (`scripts/optimize.py`, stdlib only) does it deterministically; your job is to gather the real menu/coupon/points data from the **mcd-mcp** MCP server, shape it into the solver's input, verify the top candidates' true prices, and present the plan. **This skill is an advisor by default — it does NOT place the order** unless the user explicitly asks afterward. A note on language: this SKILL.md and `references/modeling.md` are instructions for you (the model) and are written in English. Chinese appears only where it's functional — mcd-mcp's own domain terms (glossed in English) and the menu/coupon data, which comes back in Chinese. **Anything you show the END USER (the install prompt below, the final plan in Step 8) should match the language they're writing in** — Chinese for a Chinese speaker, English for an English one — not hardcoded. `