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。

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.