convez.ai
Skill
YujxZJCN

teaching-skills

YujxZJCN/teaching-skills

Ver no GitHub
BomPluginScore · de 0 a 100

Teaching Skills for Claude Code — teaching lifecycle suite for university professors

Instalação

$npx convez-ai YujxZJCN/teaching-skills

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
50
Forks
8
Acompanhando
0
Saúde
71%
Último commit
há 1 mês
Versão
v1.2.0
Issues abertas
0
Contribuidores
1

README

A tradução para o português deste README ainda está na fila.

Teaching Skills for Claude Code

English | 简体中文

A comprehensive suite of Claude Code skills for university professors, covering the full teaching lifecycle: design → build → assess → deliver → reflect → improve.

14 skills · 78 modes · 67-agent ensemble · 2 quality gates · 1 Course Passport

AI is your teaching assistant, not your replacement. This suite won't teach for you. It handles the structure and the grunt work — drafting outcomes, building blueprints before exams, auditing alignment, formatting syllabi, coding evaluation comments — so you can focus on what actually requires you: knowing your discipline, knowing your students, and deciding what matters. Every consequential decision passes through a checkpoint where you decide.

Architecturally inspired by Academic Research Skills (Cheng-I Wu) — the research-side sibling of this suite. If you also write papers, the two compose: teaching-reflector's SoTL mode hands off to ARS's deep-research and academic-paper skills.

Install

Plugin (recommended):

/plugin marketplace add YujxZJCN/teaching-skills
/plugin install teaching-skills

Manual: clone this repo and symlink the skill directories into ~/.claude/skills/ (or the project's .claude/skills/).

Verify: run /ts-status in a directory with a course_passport.yaml, or just say "Design a new course on X for 60 second-year students."

The skills

SkillStageWhat it does
course-designer1 DESIGNBackward design: Bloom-tagged outcomes → assessment plan → semester arc → syllabus. Socratic mode for blank-page starts.
lesson-builder2 BUILDLesson plans, lecture notes, slide outlines, active-learning activities, cases, discussion guides, flipped formats.
assessment-architect3 ASSESSBlueprint-first exams and quizzes, rubrics, TILT project briefs, AI-resilience integrity audits, post-exam item analysis.
student-mentor4 DELIVERFeedback writing, struggling-student intervention plans, recommendation letters, difficult emails, advising and mentoring plans.
submission-auditor4 DELIVERSpec-driven submission checking: compile your template/requirements into a checkable spec, audit submissions (single or batch) with located evidence, per-student feedback reports + class pattern report.
teaching-reflector5 REFLECTBias-honest evaluation analysis, mid-course feedback, peer observation, teaching portfolio and statement, SoTL design.
teaching-pipelineorchestratorRuns the whole lifecycle over the Course Passport, with two non-skippable gates and a weekly delivery loop.

Extension skills — production, operations, and compliance layers; each works standalone and the pipeline dispatches them on demand:

SkillStageWhat it does
deck-studio2Renders actual slide decks (Marp/Pandoc/Beamer/python-pptx), course-wide themes, code-generated figures, handouts, posters. Accessibility enforced, never fakes a render.
lab-forge2Executable STEM artifacts: lab packages, per-student synthetic datasets with recoverable ground truth, starter code, autograders, executed reference solutions.
media-scripter2Recorded-media scripting: 6–9 min mini-lecture scripts, storyboards, episode series, caption/transcript cleanup, audio adaptations.
course-publisher4Student-facing comms from the passport: announcements, weekly emails, static course site, LMS packaging, living FAQ. Drafts only — facts traced, never invented.
ta-coordinator4Teaching-team operations: TA handbooks, grading calibration sessions, hours-balanced allocation, meeting agendas, cross-TA consistency checks (no league tables).
accreditation-mapper1Outcomes → program outcomes → standards matrices with evidence status, evidence packages, gap analysis, honesty-capped self-study drafts.
bilingual-coursewaresupportTerminology-disciplined bilingual materials: professor-confirmed glossary, glossary-bound translation, paired-version sync, consistency audits.

The pipeline

Stage 0 CONTEXT  → Course Passport initialized                 🧑
Stage 1 DESIGN   → course-designer                             🧑
Gate 1.5 ALIGNMENT — constructive-alignment audit              ✓ machine + ack
Stage 2 BUILD    → lesson-builder (just-in-time by default)    🧑
Stage 3 ASSESS   → assessment-architect + integrity audit      🧑
Gate 3.5 QUALITY — AI-integrity, transparency, UDL, workload   ✓ machine + ack
Stage 4 DELIVER  → weekly loop: materials · mentoring · submission audits · midcourse feedback
Stage 5 REFLECT  → teaching-reflector eval-analysis            🧑
Stage 6 IMPROVE  → iteration record → next term re-enters Stage 1

🧑 = professor checkpoint (you decide) · ✓ = deterministic gate, then your acknowledgment

Two ideas hold this together:

  • The Course Passport (shared/course_passport_schema.md) — a YAML file that is the single source of truth for the course. Outcomes, assessment plan, schedule, policies, gate findings, artifact ledger, iteration history. State lives in the passport, not the conversation: any fresh session resumes from it. Every skill also works standalone without one.
  • Constructive alignment as a machine check — the Alignment Gate verifies the outcome–teaching–assessment triangle is closed before anything is built on it; the Quality Gate verifies the built artifacts are transparent (TILT), accessible (UDL), integrity-coherent (AI-era), and humanly workloaded — before students see them.

Quick start

# Full lifecycle
"I'm teaching a new course on environmental economics next fall — run the full pipeline"

# Blank page
"Guide me — I know my field but I've never designed a course"          → socratic

# Single artifacts
/ts-outcomes  /ts-syllabus  /ts-lesson  /ts-exam  /ts-rubric  /ts-letter  /ts-evals

# Mid-entry
"I already have a syllabus, build me week 3's materials"
"Semester's over, here are my student evaluations"                      → eval-analysis

# Status
/ts-status    "What's next for my course?"

See QUICKSTART.md for a worked first session and MODE_REGISTRY.md for all 78 modes.

Design principles

  1. The professor is the pilot. Checkpoints at every stage; "key decisions made for you" are always surfaced, never buried; "just proceed" is respected (shared/checkpoint_protocol.md).
  2. Evidence-based defaults, professor's final call. Recommendations cite shared/pedagogy_foundations.md (backward design, constructive alignment, active learning, retrieval practice, TILT, UDL, cognitive load, feedback research). When you overrule a principle, the decision is logged and never re-litigated.
  3. No invented context. Empty learner profile → the skill asks. Institutional policies → [NEEDS PROFESSOR INPUT] markers, never plausible filler. Uncertain domain claims in generated materials → [VERIFY] markers.
  4. Person-affecting outputs are evidence-bound. Feedback, letters, and intervention plans use only material you provided — no invented anecdotes, ever — and always end with a verify-before-send reminder. Student data never enters the Course Passport.
  5. Integrity by design, not detection. The AI-era integrity model (shared/ai_era_integrity.md) uses per-assignment policy tiers and structural resilience patterns. Nothing in this suite relies on AI-detection tools.
  6. Honest evidence reading. Student evaluations are treated as biased, small-sample evidence of student experience — analyzed thematically, triangulated, and caveated — not as a teaching-quality score.

Repository layout

course-designer/        SKILL.md + agents/ + references/ + templates/
lesson-builder/         〃
assessment-architect/   〃
student-mentor/         〃
submission-auditor/     〃
teaching-reflector/     〃
teaching-pipeline/      SKILL.md + agents/ + references/
deck-studio/            extension skills, same layout
lab-forge/              〃
media-scripter/         〃
course-publisher/       〃
ta-coordinator/         〃
accreditation-mapper/   〃
bilingual-courseware/   〃
shared/                 Course Passport schema · pedagogy foundations · gate protocols
                        · AI-era integrity · checkpoint protocol
commands/               /ts-* slash commands
docs/                   ARCHITECTURE.md
.claude-plugin/         plugin + marketplace manifests
skills/                 symlinks for plugin auto-discovery

Languages

English and Simplified Chinese trigger keywords ship by default; intent-based modes (socratic design, mid-entry routing) work in any language. Add your language's keywords to a skill's description frontmatter to improve trigger matching.

License

MIT. The architecture (skills/agents/gates/passport pattern) is inspired by Academic Research Skills (CC BY-NC 4.0); no content from that project is reproduced here.

Ver o original no GitHub

Bom para

Relacionados