dsh-plugin-education

When you use a large language model to author lesson plans, quizzes, or flashcards, the text appears fast, but verifying structural integrity, option validity, and readability fit still falls back on manual review. dsh-plugin-education exists to close that gap: it does not write the content itself, but equips DeepSeek Harness agents with a structured education toolkit that lets the model validate and format its own output immediately after generation.

The plugin exposes seven actions: scaffold a lesson plan with a minute budget across four grade bands, validate quiz items for stem, unique options, answer index, and explanation, build an analytic rubric from a criteria-by-levels matrix, convert Q/A pairs into Anki TSV or Markdown flashcards, assess readability via Flesch scoring for English or CJK length heuristics for Chinese, aggregate scored items into a grading sheet with totals, percentage, letter grade, and pass/fail determination, and generate a weekly study plan from a date range and per-topic hour budgets. Every piece of logic is a pure function with zero filesystem access, making it deterministic and independently unit-testable.

It suits teachers, instructional designers, and course-builders who rely on LLMs for lesson prep and assessment creation, as well as teams embedding content-QA steps into agentic workflows. Configuration is minimal, with only two optional parameters to connect.