dsh-lesson-outline

When generating teaching materials with an LLM, the most common frustration is structural drift: run the same topic three times and you get three different section splits, bullet-grain levels, and time allocations, forcing the instructor to manually reorganize every output. dsh-lesson-outline removes the outline from free-form model generation and hands it to a deterministic, pure-logic module, making the structure fully predictable.

At its core is a single lesson_outline tool: supply a topic, a section count (1-8, default 3), and a learner level (beginner / intermediate / advanced), and it returns a normalized outline JSON in which every section carries a title, key points, suggested duration, and learning activity. The model then receives this scaffold and only needs to flesh out each section into complete teaching content, rather than inventing its own chapter structure on the fly. The entire process makes no external API calls and consumes no extra tokens, running purely in the local logic layer.

Ideal for teachers, corporate trainers, and online-course creators who want a stable, reproducible course skeleton while still letting the model generate rich prose around it, saving the time previously spent reworking outlines on every run.