gaokao-pro

用 Claude Code 规划你的高考。

把下面这段 prompt 粘进 Claude Code / Codex / Cursor：

跑 `npx gaokao-pro@latest help` 把命令摸清楚，然后帮我规划 2026 年的高考志愿。

先问我：分数（估分 / 模考分 / 高考分都行，标清楚是哪种）、省份、选科组合、目标专业方向或职业兴趣、偏好（目标城市 / 是否限定 985/211 / 学费预算）。如果给的是估分或模考分，先跑 rank-tables 确认该省收录的年份科类，参考 2024-2025 历年一分一段做粗估位次并向我说明数据年份；等高考真实分数出来、且 2026 当年一分一段已收录后再精算。

每条推荐都用 CLI 拉真实数据支撑——查历年最低分、跨校搜专业、把分数换算成位次区间。

Or use the CLI yourself:

npx gaokao-pro@latest recommend \
  --score 660 --province henan --subjects 物理,化学,生物 \
  --985 --limit 5 --explain

npx gaokao-pro@latest rank --province beijing --year 2024 --score 650
# → 全省位次 3176 名以内

gaokao-pro recommend  score=660  河南(3+1+2)  subjects=物理/化学/生物  track=物理类  evaluated=44

[冲  REACH]  4 schools
  school              delta  min(2025)  city    tags  belong
  南京大学            -9     669        南京市  985   教育部
  北京航空航天大学    -10    670        北京市  985   工业和信息化部
  ...

[稳  MATCH]  4 schools
  ...

[保  SAFETY]  4 schools
  ...

No signup, no token, no backend. Talks straight to static-data.gaokao.cn (the 中国教育在线 / 掌上高考 static JSON tier — no auth, no sign, no rate limit observed). Ships with a built-in 2,400-school local index (1 MB gzipped) so recommend and top run fully offline.

Verbs

| Verb | What it does | |-------------|---------------------------------------------------------------------------| | recommend | 冲 / 稳 / 保 buckets for your score in a province (offline) | | top | Top-N best schools your score can reach (offline) | | find | Search majors across schools (e.g. all 985 schools that recruit 计算机) | | school | School metadata: 985/211/双一流/学科评估/校友会排名 | | plan | Forward-looking admission plan per (school, year, province) | | actual | Backward-looking actual admissions: 最低分 / 平均分 / 最低位次 | | scores | Historical min-score time series for a (school, province) pair | | rank | score ↔ 全省位次 via official 一分一段表 (30 省 2024+2025 双年份) | | provinces | List supported provinces with 新高考 reform regime | | mcp | Run as MCP server — claude mcp add gaokao-pro -- npx -y gaokao-pro mcp | | groups | 院校专业组 lookup (121 schools × 11 provinces × ~5450 groups × ~30,000 majors) | | xiaoce | 强基/综评 校测详情 (59 校): 笔试/面试/体测/录取分配/报名时间/签约条款 | | paths | 志愿路径全景 — 一次列出提前批/综评/运动队/省调剂rules + ✓/✗ caveat | | dossier | 院校 dossier — 7 数据集一站聚合：招生网+专业组+校测+综评+运动队+提前批+滑档 | | roadmap | 完整志愿规划 — recommend + per-pick slip-risk + paths + 关键提醒, 一次出全图 | | province-overview | 省份 overview — 调剂rules + 2026日历 + 综评+提前批+滑档历史+college覆盖 | | calendar | 2026 投档时间日历 — 31 省 batch fill→dispatch→release→supplementary 时间窗 | | huadang | 滑档/退档 历史案例 (45 case): by-province / by-category / list-categories | | slip-risk | 滑档风险评估 — (分/位次)×组×省调剂rules+组内梯度 → 4 档verdict + 中文 reasons + ≤3 案例 | | tiqian-pi | 提前批 catalog (151 项目 × 16 类型 × 38+ 省): 强基/公费师范/优师/综评/三位一体/中外合作/专项/公安/军校/港校/航海/小语种/民族班/预科 | | zongping | 综评 2026 by-school: UCAS / SUSTech / ShanghaiTech / CUHKSZ / 沪苏浙鲁粤综评校 | | gaoshui-sport | 高水平运动队 by-sport (post-2024 reform): tier × exam-window × score-path |

Run gaokao-pro help for the full command reference.

Flags

• --province <name|id> — accepts 中文 (河南), pinyin (henan), or GB id (41).
• --subjects <list> — comma-separated, e.g. 物理,化学,生物 for 3+1+2 or 3+3 provinces; subject set drives track inference (物理类/历史类/综合改革/文/理).
• --985 / --211 / --dual-class — label filters.
• --belong <教育部|工信部|...> — by 隶属.
• --limit <n> — cap results.
• --explain — show reasoning per recommendation (table mode).
• --format table|json — defaults to table on TTY, json when piped (great for jq / Claude Code).

How it scores

Recommend uses a transparent score-delta heuristic — no opaque model. For each school: delta = userScore − historicalMin(latestMatchingYear, sameTrack), bucketed as:

• 保 SAFETY: delta ≥ +15
• 稳 MATCH: −5 ≤ delta < +15
• 冲 REACH: −25 ≤ delta < −5
• out of range: delta < −25

The deltas are surfaced in every result so you can override the cutoffs.

Refreshing the index

The bundled data/school-index.json.gz was probed against static-data.gaokao.cn on 2026-05-22 (3,145 schools, includes 2023-2025 historical min scores). To rebuild against the latest upstream:

pnpm install
pnpm probe                  # ≈ 60-90s at concurrency 25, writes cli/data/school-index.json.gz

Status

Early — v0.0.2. Coming: rank-based recommendations (using min_section from schoolspecialscore), MCP server, Claude Code plugin, hosted landing at gaokao.ha7ch.com.

Part of HA7CH — sibling of job-pro and cv-pro.

License

MIT.