lingochunk-skill-author

LingoChunk skill author (the meta-skill)

Turn one finished lesson into a repeatable FORMAT. The user has typically refined a lesson by trial - co-edit rounds until the structure is exactly right - and now wants "this shape again, for any episode". That shape is a skill: a markdown playbook that any MCP-capable agent (claude.ai, Claude Code, ChatGPT dev mode, ...) can follow to rebuild the format from fresh episode content. A skill is ONLY text - the capability lives in the LingoChunk MCP tools, which are identical on every client, and the server stays the validator of record - so a skill travels without trust machinery: a bad one produces validation errors, not bad lessons.

This skill uses the lingochunk MCP tools. If they are not available, tell the user to add the LingoChunk MCP server (see the plugin README) and stop.

When to use

• "Make a skill out of this lesson so we can do the same for next week's episode."
• "Save this lesson format as a template."
• "I want my students' agents to be able to build lessons in this format."
• "Turn the way we structured this into something reusable."

Not for building a lesson (that is lingochunk-lesson or a skill you already made), and not for one-off edits to an existing lesson (update_lesson directly).

Options to settle first (ask only what the user left open)

1. Source lesson: which saved lesson embodies the format (list_lessons if they don't say; usually the one just co-edited).
2. How strict a clone: exact format (same spine, same exercise types, same counts) or a genre (the spine and voice fixed, the exercise mix allowed to vary by episode/level - say which parts flex).
3. Where it lives: private (the user keeps the markdown; default) or contributed to the public lingochunk-mcp repo for everyone. Never assume publication.

Workflow

1. Fetch and read the lesson. get_lesson -> document. Extract the format signature:
   • the spine: the section sequence and what sits under each;
   • the exercise mix: which blocks, how many items, where anchors and audio windows are used;
   • the voice: instruction style and language, prose tone, dialogue note habits, show_sentence usage, highlight policy, review-block shape;
   • the policies implied: level pitch, which FSRS states get drilled, glossary size.