DSH HUB
HomePlugin StorePlugin PacksCommunityRankingsResourcesPublish Guide
Plugin source
Back to catalog

XSY-28 /

XSY-28/learning-skills

Verified

Learning skills for Codex and DeepSeek Harness: course notes, conceptual explanations, and self-contained lectures.

★ 3 Stars0 Forks0 IssuesN/A Community rating0 Confirmed installs
View on GitHub
READMESource: main@1f09cee2

Learning Skills

中文

Four skills for Codex and DeepSeek Harness: turn lectures into notes, work through confusing concepts, write self-contained chapters, and understand complex systems through their evolution.

Each skill is a folder of agent instructions. course-to-notes also includes scripts for subtitles, video frames, and output checks. You can install any of them on its own. The instructions are mainly in Chinese; you choose the output language.

Choose a skill

What you want to do Skill
Find a course that fits your background, or turn a lecture into notes with timestamps and relevant images course-to-notes
Understand a project, theory, or institution through its key changes and design tradeoffs How an Aircraft Carrier Is Built
Understand why a definition, method, or step in a proof makes sense explain-for-understanding
Write or substantially revise a lecture chapter, including the background needed to follow it obsidian-lecture-writing

Install

DeepSeek Harness

Install all four as a DSH plugin in your chosen profile:

dsh plugin --profile web add github:XSY-28/learning-skills

In DSH, invoke /course-to-notes, /evolutionary-understanding, /explain-for-understanding, or /obsidian-lecture-writing. See the DSH guide for prerequisites, examples, and updates.

Codex

In Codex, send:

Use $skill-installer to install these paths from XSY-28/learning-skills:
- skills/course-to-notes
- skills/evolutionary-understanding
- skills/explain-for-understanding
- skills/obsidian-lecture-writing

Keep only the paths you want. After installation, send a new message using one of the prompts below. If Codex cannot find the installed skill, restart it.

The installer refuses to overwrite an existing skill folder. Back up any local changes before replacing an older copy.

Install from the terminal or copy the folders manually

If your Codex installation has the bundled installer at the standard location:

python3 "${CODEX_HOME:-$HOME/.codex}/skills/.system/skill-installer/scripts/install-skill-from-github.py" \
  --repo XSY-28/learning-skills \
  --path skills/course-to-notes skills/evolutionary-understanding skills/explain-for-understanding skills/obsidian-lecture-writing

Or download this repository and copy the skill folders you want from skills/ into your configured skills directory. Copy each whole folder, including its references, scripts, metadata, and license.

Use

The examples below use Codex syntax. In DSH, replace $skill-name with /skill-name, as shown in the DSH guide.

How an Aircraft Carrier Is Built

The Chinese display name is “航母是怎样建成的”; the invocation identifier remains evolutionary-understanding. The skill applies to the evolution of complex systems across domains. Both the instructions and software-history guide are available in English and Chinese through their language links.

Use $evolutionary-understanding to explain the major design changes in DSH.
Use its Git history, connect each change to its constraints and tradeoffs,
and distinguish documented history from inference and teaching examples.

The skill selects changes by their effect on behavior and structure, preserves branches and reversals, and relates them to the current system. Section titles name the concrete change directly. It also applies to theories and institutions; it does not promise faster learning or treat a coherent narrative as evidence of mastery.

Work through a concept

Use $explain-for-understanding to explain regularization.
I can take derivatives, but I don't understand why we add a penalty to the loss.
Start with a small example I can calculate by hand.

The skill asks Codex to explain what each quantity represents, connect the example to the derivation, and state when the conclusion applies.

Turn a lecture into notes

Use $course-to-notes to turn the first 10 minutes of the lecture I provide
into Chinese Markdown notes in ./course-notes.

The workflow preserves the lecture's teaching sequence, derivations, questions, and meaningful repetition. Notes include source timestamps and relevant images where the material is available; missing or unclear material is reported. You can supply learning records to get chapter-level reading suggestions; without records, prior learning stays unconfirmed.

It reads the full available transcript first, maps the teaching sequence, then checks visuals at relevant source timestamps. Necessary playback can use a supported faster speed, with pauses or slower review for dense derivations and demonstrations. Transcript gaps and unspoken board work still need checking; fewer screenshots do not establish complete coverage. This changes the processing workflow, without guaranteeing a fixed speedup.

If you have not chosen a course, give your background and goal and ask for course recommendations first.

Write a chapter

Use $obsidian-lecture-writing in detailed mode to write an introductory chapter on conditional
expectation for someone who knows basic probability. Deliver it in this chat.

The skill asks Codex to introduce concepts before relying on them, explain the reasons behind key steps, and review the chapter from a reader's position. To save it in Obsidian, specify your vault and destination file.

Choose simplify for a compact, outline-style chapter or systematic review, and detailed for step-by-step explanations, worked examples, and necessary prerequisites. Both preserve key relationships and conditions. The mode controls explanation detail, not theoretical depth or topic coverage; individual difficult sections can be expanded without changing the whole chapter. If no mode is specified, the skill uses the current learning purpose and asks only when the distinction materially affects the result. Syntax and formula lookups remain outside this skill in either mode.

These workflows are for learning and teaching. A formula lookup, translation, or formatting fix can stay brief. Automatic skill selection depends on the client; use $skill-name when you want to choose explicitly.

Requirements

The explanation, chapter-writing, and evolutionary-understanding skills have no script dependencies. The course helpers use Python 3.10+ and the standard library; extracting video frames also needs ffmpeg and ffprobe on PATH.

Video access, audio transcription, and Word/PDF export depend on tools available in your agent environment. This repository does not include a video downloader, transcription model, or document converter.

Checks and contributions

See validation.md for development commands, test coverage, and what has not been verified. Package and helper-script checks do not establish that generated notes are correct or that a learner has mastered the material.

For a bug report or contribution, include a concrete request, what happened, and what you expected. Keep private notes, learning records, credentials, and third-party course media out of commits.

MIT licensed. The license covers this repository's original instructions, code, and test fixtures; course materials retain their own terms.

—/ 5

No ratings yet

Verified DSH bundle

Commit 1f09cee2da79

Community comments

No comments yet. Be the first to write one.

DSH HUB

A community index for DSH plugins. Not an official GitHub or DeepSeek AI product.

CommunityResourcesAPIAbout