Coach

Coach helps you write course lesson content with an AI agent and shapes it to Andamio's course model, ready to publish through the CLI.

Coach is a public Agent Skills package for authoring courses with an AI agent. Its job is simple to state: it helps you write lesson content and matches it to Andamio's model, so what you author with an agent publishes cleanly onto the platform through the Andamio CLI.

Full documentation

Coach is open source, and the coach repo is the source of truth for the complete skill set and workflow. It is also published on npm as @andamio/coach. The rest of this page is a quick orientation.

From lesson content to Andamio

You write the lessons with an agent; Coach structures them the way Andamio expects, at every step:

Learning targets. Coach drafts and assesses Student Learning Targets, the "I can..." statements that define each module's outcome. One SLT per module.
Lessons. It classifies each SLT into a lesson type, helps you gather the context that type needs (screenshots, code examples), and writes the lesson alongside you.
Compile. The /compile skill packages finished modules into Andamio's import format: outline.md, numbered lesson files, and optional introduction.md / assignment.md.
Publish. Hand the compiled module to andamio course import, and it lands on the platform, ready to register and publish on-chain.

That is the "matches it to Andamio's APIs" part: Coach produces exactly the structure the CLI and the platform consume, so there is no manual translation between authoring and publishing.

Three ways to start

Run the start skill and pick a path. Three entry points, one destination.

Path	Use when
Beginner	You want to design a learning experience about something new. Conversation-first.
Apprentice	You know some of the content and are ready to build a course. Collaborative drafting.
Teacher	You have learning targets ready and are an expert on the subject. Fast-track setup.

Install

Coach follows the open Agent Skills standard, so it works with any compatible agent (Claude Code, Cursor, GitHub Copilot / VS Code, Gemini CLI, OpenAI Codex, Goose, Roo Code, and more). Point your agent at the skills/ directory.

# Claude Code (plugin)
/plugin marketplace add Andamio-Platform/andamio-marketplace
/plugin install coach@andamio

# npm
npm install -D @andamio/coach

# Any other agent (clone, then point it at skills/)
git clone https://github.com/Andamio-Platform/coach.git

The compounding loop

Coach is built so each course you develop makes the next one faster. The course-workflow skill orchestrates the loop and tracks which phase you are in.

Skill	What it does
`draft-slts`	Generate learning targets from topic, audience, and goals
`assess-slts`	Evaluate SLT quality across five dimensions, suggest rewrites
`classify-lesson-types`	Interview to assign each SLT a lesson type
`self-assess-readiness`	Rate the agent's coaching confidence per SLT
`gather-screenshots` / `gather-code-examples`	Build context checklists for Product Demo and Developer Documentation lessons
`compile`	Package modules into Andamio's import format
`compound`	Extract patterns into `knowledge/` so future runs start smarter

Lesson types

Every SLT is authored as one of five lesson types, each with its own inputs:

Product Demo: SLT plus screenshots, a visual walkthrough.
Developer Documentation: SLT plus code and a docs link, a technical lesson.
How To Guide: SLT plus optional materials, a step-by-step procedure.
Organization Onboarding: SLT plus org context, setup guidance.
Exploration: SLT plus framing questions, for big ideas and worldview.

Key concepts

Student Learning Targets (SLTs): measurable "I can..." outcomes, one per module. Coach helps you draft, assess, and refine them.
Fist to Five: a 0 to 5 confidence rating (adapted from Marzano) the agent emits per SLT across four dimensions, conceptual explanation, code demonstration, learner assessment, and knowledge currency.
The skill is the student: Coach flips the classroom question. Instead of "can this student do these things?", it asks "given the context, can these agents do these things?"