作业批改与学生学业综合评估Skill

中国中小学作业批改与学生学业综合评估。教师拍照扫描批改作业、生成单生/全班单科及综合学科知识掌握评估图并给出指导建议。触发场景：(1)教师上传/拍照学生作业进行批改 (2)查询单个学生或全班的知识掌握情况 (3)生成单科或综合学科评估报告 (4)按角色（班主任/单科老师/校级管理层）分级查看学习数据。

Security Scan

The skill's instructions, references, and scope are coherent with a homework‑grading and student assessment tool; it is instruction-only, requests no unrelated credentials, and mainly recommends local OCR deployment — but operators must handle student personal data carefully and secure any deployed services.

Purpose & Capability

The name/description (homework grading and student assessment) matches the included SKILL.md and reference docs: OCR selection/implementation, data model, chart specs and permission model. Required items (none) are proportionate — there are no unrelated env vars, binaries, or config paths requested.

Instruction Scope

Instructions focus on OCR-based grading, generating charts, and role-based reports. The SKILL.md explicitly recommends local deployment (PaddleOCR Docker) and mentions cloud OCR as alternatives, and it preserves teacher human review for subjective answers. Note: using cloud OCR or model micro‑tuning would require sending student work (PII) offsite or collecting training data — the doc calls this out and recommends desensitization, but operational safeguards are needed.

Install Mechanism

This is an instruction-only skill with no install spec. It references publicly available artifacts (docker image paddlepaddle/paddleocr:latest) and standard tooling (Docker, curl). No downloads from obscure URLs or embedded installers are present in the skill bundle.

Credentials

The skill declares no required environment variables or credentials. The documentation does describe optional cloud OCR providers (Baidu/Ali/Tencent) which in real deployments would require API keys and possibly account-level credentials; the skill does not request those explicitly, which is reasonable for an instruction-only design but operators must supply and protect such credentials themselves if they choose cloud options.

Persistence & Privilege

The skill is not always-enabled and is user-invocable; it does not request persistent platform privileges. The permission model and role matrix are documented; there is no instruction to modify other skills or system-wide agent settings.

Assessment

This skill is internally consistent and appears to be a design/spec for a grading system rather than an executable package. Before adopting it, consider: 1) Data privacy — student work contains personal data; prefer local/on-prem PaddleOCR deployment or ensure you desensitize names/IDs and review cloud provider contracts and compliance (PIPL, local education regulations). 2) Network exposure — the example docker run exposes port 8866; avoid binding to public interfaces and restrict access via firewall or reverse proxy. 3) Model fine‑tuning — collecting student handwriting for micro‑tuning can create sensitive training datasets; obtain consent and secure storage. 4) Credentials — if you use cloud OCR, create least‑privilege API keys and rotate them; the skill does not request keys, so you will need to supply them safely. 5) Verify provenance — source/homepage is unknown and the file claims Huimai Intelligence copyright; if you plan commercial use or production deployment, obtain the appropriate license/agreements and request source code or a vetted distribution. 6) Operational testing — validate OCR accuracy and keep teacher review in the loop (the docs recommend this). If you want a higher assurance assessment, ask the publisher for: an official homepage/repository, a signed distribution or container image provenance, and any sample deployments or SOC/compliance docs.

作业批改与学生学业综合评估

核心功能

1. 作业批改（拍照/扫描输入）

• 输入方式: 教师通过手机拍照或扫描学生作业（纸质或电子版）
• OCR+AI 识别: 自动识别学生作答内容，并与标准答案/知识点映射对照
• 批改结果:
  • 逐题标注对/错，标注错误类型（概念错误、计算错误、审题不清、表述不规范等）
  • 每题关联对应的课程标准知识点
  • 生成错题提示（告诉学生错在哪、正确思路是什么）
  • 支持主观题人工复核接口

2. 知识点掌握评估图

• 单科维度（单学生/全班）:
  • 雷达图/柱状图：展示各知识点掌握率（0-100%）
  • 知识点分级：掌握/需巩固/薄弱 三级
  • 趋势线：近期历次测验的知识点掌握变化
• 综合学科维度（单学生/全班）:
  • 学科间对比雷达图（语文/数学/英语/科学等）
  • 各学科综合素养评分
  • 学科均衡度分析

3. 指导建议生成

• 单学生层面:
  • 薄弱知识点专项练习题推荐
  • 个性化学习计划建议
  • 家长端简要报告（适合家长理解的语言）
• 全班层面:
  • 班级共性薄弱知识点汇总
  • 教学调整建议（哪些知识点需要重新讲授、哪些学生需要重点关注）
  • 分组教学/分层作业建议

4. 分级权限体系

角色	可见范围
单科教师	所教班级单科知识掌握情况、班级内学生个人单科情况
班主任	所带班级全学科综合情况、班级内学生个人全科综合情况
年级组长	年级内各班各科横向对比、年级整体水平报告
校级管理层	全校各年级学科分析报告、教师教学质量综合数据

数据结构参考

见 references/data-model.md — 学生、班级、学科、知识点、作业记录数据模型。 见 references/permission-model.md — 分级角色权限映射。 见 references/chart-spec.md — 评估图表规格说明。 见 references/ocr-implementation.md — OCR 引擎选型（PaddleOCR推荐）、拍照批改工作流、Docker部署方案。 见 references/textbook-version-mapping.md — 国内教材版本、三层知识点映射、5张数据表设计、跨版本兼容方案。

---

授权说明

版权与知识产权: © 2026 Huimai Intelligence. All Rights Reserved.

本技能文件（含所有附属参考文档）为设计说明与产品预览，仅限个人教师免费查阅和使用其中的教育理念与评估方法论。

商业使用许可: 任何学校、教育机构或企业如需部署、集成、定制本系统，或进行任何形式的商业使用，须与惠迈智能（Huimai Intelligence）签订书面商业许可协议。

本文件发布于 ClawHub 平台，发布时间戳作为原创性证明，受《中华人民共和国著作权法》及相关国际版权条约保护。未经授权复制、分发、修改或商用将追究法律责任。