@hobin/learning

English | 한국어

Five independent Pi Skills for understanding technical sources, studying public repositories, forming reusable ideas, and practicing them.

Learning is not a course runner or a fixed pipeline. Each Skill can be used by itself and ends with its own complete result.

Install

Requires Pi and Node.js 22.19 or newer.

pi install npm:@hobin/learning

Try it for one run:

pi -e npm:@hobin/learning

Try this first

Ask Pi normally:

Read this RFC section with me. Preserve the example and the exception, then
explain the state model it expects me to use.

Or invoke a Skill directly:

/skill:technical-reading Explain this article without flattening its boundaries.
/skill:opensource-reading Trace this API through docs, tests, and implementation.
/skill:conceptualize Turn these findings into one source-independent concept.
/skill:patternize Decide whether these repeated cases form one reusable routine.
/skill:exercise Build prediction, diagnosis, repair, and transfer practice.

/learning opens a small chooser. It places the selected /skill:... command in the editor and preserves the current draft. It does not send the command or start a separate Learning session.

Choose a Skill

Needed result	Skill
Understand a book, article, specification, tutorial, PDF, or webpage faithfully	technical-reading
Learn how one public API, flow, invariant, or tradeoff works in an open-source repository	opensource-reading
Name and test a durable idea that survives beyond its sources	conceptualize
Turn recurrence across cases into an operational routine	patternize
Produce observable evidence that a learner can use the idea	exercise

These are different questions, not five phases. Technical reading does not have to become a concept; a concept does not have to become a pattern; any completed result can stop where it is.

What happens after selecting a Skill

1. Pi loads the selected SKILL.md as the complete method.
2. The Skill reads only the source or repository slice needed for its question.
3. A packaged reference is opened only when its stated distinction can change the result.
4. The Skill returns a complete conversational result or copyable Markdown.
5. It writes a file only after the user names or approves a target.

Learning does not create a notebook, graph database, progress record, or hidden cross-Skill state.

Optional reference directions

Each Skill has a small judgment.json used during package development. It is rendered into a deterministic Context Directions section inside SKILL.md. The installed package does not start a Judgment runtime. Pi simply shows the method and directions to the model as normal Skill content.

Documentation

• Choosing a Skill — the distinction between the five results
• Learning operating principles — Pi resource discovery, editor staging, prompt-time Skill execution, and the authoring/runtime boundary

Development

pnpm --filter @hobin/learning check
pnpm --filter @hobin/learning eval
pi -e ./packages/learning

After changing a policy, regenerate the embedded directions:

node packages/learning/scripts/write-context-directions.mjs

License

MIT