jcc-s17-lineup

name	jcc-s17-lineup
description	Use when answering questions about 金铲铲 S17棋子、羁绊、羁绊档位、阵容筛选、装备、海克斯、经济运营或根据自然语言条件组建阵容. Includes bundled data and Python tools for fast unit/trait lookup, knowledge search, lineup building, and trait evaluation.

金铲铲 S17 阵容助手

Use this skill for 金铲铲/云顶类 S17 questions involving units, traits, trait breakpoints, items, augments, economy notes, or lineup composition.

Data

Bundled data lives at assets/s17_data.json and is generated from:

• assets/source-workbooks/金铲铲S17棋子羁绊大全.xlsx
• assets/source-workbooks/金铲铲S17赛季完全辅助手册_v2.xlsx

The merged dataset includes:

• season_overview
• units
• traits
• items.base_items
• items.completed_items
• items.emblems
• augments
• economy.leveling_table
• economy.shop_odds
• economy.income_rules
• economy.stage_guidance
• economy.tips

Trait breakpoints are parsed from trait effect text:

• (2), （3）, (4) style markers are treated as unlock counts.
• Text like 2未来 / 3未来 / 4未来 is also parsed for 未来战士.
• If a trait has no explicit breakpoint text, the tools treat 1 as active so unique or descriptive traits still count.

Tools

Run scripts from the skill directory or pass --data assets/s17_data.json.

python3 scripts/query_units.py 游侠
python3 scripts/query_units.py --trait 暗星 --cost 4
python3 scripts/query_traits.py 暗星
python3 scripts/query_knowledge.py 白银 --category augments
python3 scripts/query_knowledge.py 4费 --category shop_odds
python3 scripts/query_knowledge.py 蓝盾 --category completed_items
python3 scripts/build_lineup.py "在8人口下，开启羁绊数量最多的8个棋子"
python3 scripts/build_lineup.py "在7人口下，优先开启暗星和游侠"
python3 scripts/evaluate_lineup.py 贝蕾亚 雷克塞 卑尔维斯 卡莎 菲兹 泰隆 锐雯 烬

Workflow

1. For unit lookup, call query_units.py.
2. For trait lookup and breakpoint questions, call query_traits.py.
3. For items, augments, season notes, leveling, D牌概率, or运营建议, call query_knowledge.py.
4. For lineup requests, call build_lineup.py with the user's original Chinese instruction.
5. For a user-provided lineup, call evaluate_lineup.py with unit names.
6. Summarize the returned JSON in Chinese with the fields most relevant to the user question.

Lineup Scoring

build_lineup.py uses beam search plus local swaps. The default objective prioritizes:

1. number of active traits,
2. explicitly requested traits and units,
3. higher active breakpoints,
4. closeness to next breakpoints,
5. total unit cost as a final tie-breaker.

This is a deterministic recommendation helper, not a live meta-tier ranking. State that limitation when the user asks for strength, meta, or win-rate.

Refreshing Data

If the Excel file changes:

python3 scripts/extract_data.py

By default, the script reads the workbooks bundled inside assets/source-workbooks/. Then rerun the relevant query or lineup command.