agent-skills-for-students

Agent Skills for University Students

English | Chinese

Skills for AI agents that follow the SKILL.md convention: a directory, a YAML frontmatter block, and a procedure the agent loads only when the trigger fires. Nothing here depends on a particular vendor or client, so the same tree works in any agent that reads this format.

78 skills, grouped by what they are for. Two of them carry the rest.

The two that matter

knowledge-acquisition-protocol

A six-stage loop for learning one unit of material to the point where you can teach it: capture > comprehend > recall > apply > teach > archive. Each stage names an action, an artefact and a pass condition, so "I studied it" becomes something checkable.

Four laws hold the loop together: think for fifteen minutes before asking; the agent guides and never hands over the answer; explaining is the only pass mark; the agent's job is to find holes in your work.

The useful part is the failure handling. When you are wrong, the agent does not correct you: it follows your premise until something absurd falls out, then asks where the argument broke, so you find the faulty assumption yourself. Recall happens in conversation and leaves no file. Every error has to compress into one stated rule before the stage closes.

vault-integration-standard

The schema that keeps a course-based notes vault navigable: a course tree for order, wikilinks for reach, frontmatter for identity.

Three navigation layers (vault > semester > course hub), a frontmatter contract every note shares, and one footer block of three link classes. Any note is reachable in four steps. Integration merges missing keys and never overwrites, never deletes, so the pass is auditable and reversible.

It also explains why: which knowledge-management functions each mechanism covers, why atomic notes with explicit links suit course-based material, and what the design deliberately refuses to do.

Install

git clone https://github.com/ZileiC/agent-skills-for-students.git

Then copy what you want into the directory your agent scans for skills:

cp -r agent-skills-for-students/skills/* <your-agent-skills-dir>/

Most agents load skills at start-up. INSTALL.md covers installing a single skill, where the directory usually lives, and how to check that a skill registered.

How a skill is shaped

skills/<group>/<skill-name>/SKILL.md

The frontmatter carries name and description; the description is the only part the agent reads before deciding to load the body. The body answers four questions: when to use it, what the procedure is, where it goes wrong, and how to verify the result.

Supporting material sits beside it in references/, templates/, scripts/ or assets/.

What is inside

Agent engineering

Delegation, review, debugging and delivery discipline for agent work.

Teaching and study

Tutoring, grading, exam preparation and the study loop.

Knowledge base

A vault that stays navigable: protocol, schema, retrieval.

Decks and documents

Stage decks, PDF and PPTX, typography, delivery copy.

Research and media

Search, citations, papers, transcripts, generated media.

Operations

Files, mail, OCR, monitoring and the small mechanics of a working desktop.

Licence

MIT. See LICENSE.