By youseiushida
Tutoring skills (project-init, tutoring, treatment-shift, graph-edit, session-wrap) for problem-driven learning. Requires the benkyo CLI to be installed separately (uv tool install benkyo).
Set up a new benkyo learning project, or resume a project after a long gap. Use this skill whenever the learner says they want to start studying a subject ("○○を勉強したい", "○○の試験勉強", "○○ができるようになりたい"), shares course materials (textbook, lecture recordings, past exams, syllabus), expresses a learning goal without yet having a benkyo project, or returns to an existing project after a long break. Trigger when no benkyo project exists for the topic, OR when the learner needs help defining what to learn. Also trigger when the learner is goal-less ("何か勉強したい") and needs help formulating a concrete target.
Conduct learning sessions using the benkyo CLI. Use this skill whenever the learner is actively working on a problem, asking to be taught a concept, attempting a probe, saying things like 「分からない」「教えて」「問題やる」「次なに」「分かった」「テストして」「ちゃんと理解したい」「公式覚えれば」, or otherwise engaged in real-time tutoring within an existing benkyo project. Trigger even when "tutoring" is not stated explicitly — if a learner is mid-session with a benkyo project, this skill should be active. Also trigger when the learner is stuck mid-solving, when they've just attempted a problem, when they want a treatment shift (deeper understanding or shortcut to memorization), or when they need to verify their own understanding.
Change the depth-of-engagement (blackbox vs whitebox treatment) for a concept in a benkyo project. Use this skill when the learner expresses 「ちゃんと理解したい」「腹落ちしたい」「導出やりたい」「なぜそうなる?」「もっと深く」 (commit signals — wants deeper understanding), or 「公式覚えれば」「暗記でいい」「ざっくり」「飛ばして」「もういい」 (release signals — wants to shortcut to memorization). Also trigger when you (the tutor) detect indirect signals: repeated "なんで" on a concept, mis-applying a blackbox reference, failed transfer to varied problems (commit candidates), or stuck-multiple-breakdowns / fatigue / time-pressure (release candidates).
Edit the structure of a benkyo concept/problem graph — add or remove nodes and edges, adjust granularity, fix duplicates. Use this skill when the learner says 「これも追加して」「これとあれは関連してる」「これは別物」「これも勉強したい(新概念)」「これとこれ同じ」「これは分けるべき」, or when you (the tutor) notice during a session that a concept the learner just mentioned isn't in the graph yet, or that a relationship between existing nodes is missing/wrong. Also trigger when reviewing or refining the graph structurally — e.g., during project init when assessing the initial set of nodes.
Wrap up a benkyo learning session cleanly and prepare for resumption. Use this skill when the learner says 「今日はここまで」「終わり」「また明日」「疲れた」「時間ない」「もう寝る」, or signals they're about to stop. Also trigger when you (the tutor) notice the natural completion of a topic and want to propose a stopping point, or when a session has gone long enough that a wrap is appropriate. This skill handles the recap, delayed JOL seeding, and persistence of state into project.metadata so the next session can resume cleanly.
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
A research-grounded tutor that defers to you, not to its training data.
A Python CLI plus a SKILL.md bundle that turn an AI coding agent — Claude Code today, OpenAI Codex CLI with one symlink, anything else that consumes the open Agent Skills format — into a tutor with persistent memory. The agent tracks, per project, which concepts you want to truly understand and which you've decided to just use as a tool. Each operational rule cites the cognitive science it's built on.
Status: β. CLI in English. SKILL.md files are in English with Japanese-first natural-language examples (the agent adapts to the learner's language at runtime; English / other-language end-to-end use works in principle but is not yet evaluated). First-class plugin support is for Claude Code; Codex CLI works via manual install (see below).
Terminology. A project in benkyo is a learning unit — one subject (e.g., Laplace transforms) bundled with the goal problems that anchor it and the per-concept treatment decisions for that subject. Each concept has a per-project treatment: whitebox (understand the why, including derivations) or blackbox (use as a tool, formula in / answer out). The cardinal-vocabulary rule in the skills (see
.claude/skills/) translates these into the learner's natural language — the internal terms never appear in tutor speech. The current example translations are Japanese-first.Note on the math-education literature: papers we cite (Sinha & Kapur 2021; Hiebert & Lefevre 1986) call these conceptual and procedural knowledge respectively. benkyo renamed in v0.3.0 to avoid the unrelated ACT-R use of "procedural" (= automated expert knowledge), which is the opposite axis from what those papers mean. The mapping is one-to-one: paper's "conceptual" = benkyo's "whitebox"; paper's "procedural" = benkyo's "blackbox".
The Python CLI is the same for every agent:
uv tool install benkyo # or: pipx install benkyo
benkyo --version # confirm it's on PATH
Then install the skills for your agent of choice:
Claude Code (first-class plugin support):
/plugin marketplace add youseiushida/benkyo
/plugin install benkyo
Restart Claude Code; the 5 skills appear in /help.
OpenAI Codex CLI (first-class plugin marketplace; the repo ships .codex-plugin/plugin.json + .agents/plugins/marketplace.json):
codex plugin marketplace add youseiushida/benkyo
# Then in the Codex TUI: open the plugin directory, find "benkyo", install.
The same SKILL.md files are picked up by both agents (the repo carries .claude-plugin/marketplace.json for Claude Code and .codex-plugin/plugin.json for Codex, both pointing at the shared .claude/skills/ tree — they coexist without conflict). Codex's central Plugin Directory listing is coming soon; until then, codex plugin marketplace add from this repo is the supported flow.
Other SKILL.md-compatible agents (Cursor, VS Code Copilot, Gemini CLI, ...): the frontmatter and body are agent-neutral. Point your agent's skill loader at .claude/skills/benkyo-* (or copy/symlink them into its skills directory). The bundle uses the open Agent Skills format.
Drop your study materials — past exams, the textbook PDF, the syllabus, lecture notes — into the directory you launch Claude Code from. Then just describe what you want:
You: I have the past 5 years of finals for ECE 220 (signals & systems), the
textbook PDF, and the syllabus. The exam is in 12 days. Help me prep.
benkyo-project-init reads the materials, extracts the concept set and dependencies from how the textbook structures them, turns past-exam problems into the project's goal problems, proposes which concepts to treat as "use the formula" vs "really understand," asks you to confirm, and hands off to benkyo-tutoring for the first activity.
This works for any STEM domain where the curriculum is structured and past problems exist: classical mechanics, organic chemistry, algorithms, statistics, fluid dynamics, ML theory, and so on.
Before you commit time, or whenever you want to step back during a session, ask Claude:
You: Show me the map. / 全体見せて
You: What does the plan look like?
You: Where am I in this?
npx claudepluginhub youseiushida/benkyo --plugin benkyoUltra-compressed communication mode. Cuts ~75% of tokens while keeping full technical accuracy by speaking like a caveman.
Frontend design skill for UI/UX implementation
Comprehensive UI/UX design plugin for mobile (iOS, Android, React Native) and web applications with design systems, accessibility, and modern patterns
Memory compression system for Claude Code - persist context across sessions
Marketing skills for AI agents — conversion optimization, copywriting, SEO, paid ads, ad creative, and growth
Standalone image generation plugin using Nano Banana MCP server. Generates and edits images, icons, diagrams, patterns, and visual assets via Gemini image models. No Gemini CLI dependency required.