studykit

Two Claude Code skills for structured self-study with adaptive spaced repetition.

The problem

Self-study breaks down in predictable ways:

• No structure — "learn DP this week" isn't a plan. No progression, no checkpoints, no idea what to cut when you're behind.
• No retention — study something Monday, forget it by Friday. No systematic review.
• No accountability — skip a day, then two, then quit. Nobody tracking whether you actually showed up.
• For interviews: no strategy — grinding LC randomly is slow. Every company tests differently. Memorising solutions doesn't transfer.
• AI does it for you — ask Claude for help with an exercise and it writes the whole solution. You haven't learned anything.

How studykit works

Two skills that coordinate through shared data:

/study-plan — builds a structured study plan through interview + research. Classifies what you're preparing for (interview, exam, or general learning), identifies the target, researches it in the background, then builds a day-by-day schedule with topic progression, SR card integration, diagnostic checkpoints, and triage strategy.

/study-session — runs daily sessions adapted to where you are. Loads your plan, checks what's due, teaches new material, generates exercises, weaves review cards throughout, tracks everything (timing, accuracy, progress), and commits to git.

What it does differently

Problem	Solution
Forget what you learned	SM-2 spaced repetition — cards woven into session transitions, not a separate review block
No exercise discipline	Learning output mode — Claude creates exercises but refuses to write code during the solve phase
Generic interview prep	Background research on your specific company/role; pattern-based progression (15 core patterns, not random problems)
Solutions don't transfer	Primitive extraction — after solving, decomposes into transferable building blocks and generates SR cards for those
Skip days, lose track	Missed-day handling, schedule adaptation, honest progress tracking

Modes

• Technical (coding, algorithms): file-based exercises with test cases, language-aware
• Knowledge (law, history, theory): JSON-backed spaced repetition, interstitial card review
• Mixed: both in one project

Installation

git clone git@github.com:ettrickshepherd/studykit.git ~/dev/studykit

ln -sfn ~/dev/studykit/study-plan ~/.claude/skills/study-plan
ln -sfn ~/dev/studykit/study-session ~/.claude/skills/study-session

Verify:

ls -la ~/.claude/skills/study-plan/SKILL.md
ls -la ~/.claude/skills/study-session/SKILL.md

Prerequisites: Python 3.10+, uv. For coding exercises, you need the runtime for your language (Python comes with uv; JS/TS needs Bun).

Recommended for coding sessions:

/output-style learning

Architecture

/study-plan                          /study-session
    │                                     │
    ├── SKILL.md (workflow)               ├── SKILL.md (workflow)
    ├── references/                       ├── references/
    │   ├── interview-guide.md            │   ├── session-workflow.md
    │   ├── plan-templates.md             │   ├── sr-queries.md
    │   ├── sr-schema.md                  │   └── exercise-patterns.md
    │   ├── user-profile.md (created)     └── scripts/
    │   └── plans/                            ├── sr_review.py
    │       ├── _index.json                   └── session_summary.py
    │       └── <project-slug>.md
    └── scripts/
        ├── json_helpers.py (shared)
        └── init_study_project.py

How they coordinate: /study-plan writes plans to study-plan/references/plans/ with YAML frontmatter and registers them in _index.json. /study-session reads the index to discover plans, loads the selected plan's frontmatter for the project path, and connects to the project's data/ directory.

Study project directory (created per project)

~/study/<project-name>/
├── plan.md                  # The commitment — goals, strategies, what was agreed to (immutable)
├── learner-context.md       # Project-specific learner observations (updated after notable sessions)
├── learning-schedule.md     # Living day-by-day schedule (updated every session)
├── progress-report.md       # Running status, session log table, topic mastery
├── data/
│   ├── cards.json           # SR card deck — SM-2 state, review history per card
│   ├── sessions.json        # Session log — timing, topics, cards, exercises
│   ├── exercises.json       # Exercise tracking — attempts, times, outcomes
│   └── topics.json          # Topic mastery and metadata
├── daily-notes/
│   └── DD-MM-YYYY.md        # Pre-session + per-task + after-session notes
├── exercises/               # Technical mode: exercise files by topic
│   └── <topic>/
│       └── <exercise>.py
└── materials/               # User-provided reference materials

Scripts

All scripts are zero-dependency (Python stdlib). Both skills share json_helpers.py for data mutations.

json_helpers.py — Shared data operations

uv run python3 ~/.claude/skills/study-plan/scripts/json_helpers.py <command> <args>

Command	Description
due-cards <cards.json>	Cards due today, sorted by overdue-first then lowest ease
add-card <cards.json> '<json>'	Append a card with auto-ID and SM-2 defaults
update-card <cards.json> <id> '<json>'	Apply SM-2 update after review
add-session <sessions.json> '<json>'	Log a session
add-exercise <exercises.json> '<json>'	Log an exercise
stats <cards.json>	Card statistics (total, due, mature, accuracy)
progress <cards.json>	Per-deck breakdown (total, due, mature, struggling, new)
sm2 <quality> <ef> <interval> <reps>	Standalone SM-2 calculation

init_study_project.py — Project scaffolding

uv run python3 ~/.claude/skills/study-plan/scripts/init_study_project.py \
  --name <slug> --location <path> --mode <technical|knowledge|mixed> \
  [--language python] [--topics "arrays,hashing,dp"]

sr_review.py — Session review helper

uv run python3 ~/.claude/skills/study-session/scripts/sr_review.py due <cards.json>
uv run python3 ~/.claude/skills/study-session/scripts/sr_review.py summary <cards.json>

session_summary.py — Session statistics

uv run python3 ~/.claude/skills/study-session/scripts/session_summary.py brief <project-dir>
uv run python3 ~/.claude/skills/study-session/scripts/session_summary.py streak <sessions.json>

Design Decisions

Decision	Choice	Why
Data storage	JSON files	Zero deps, git-friendly diffs, Claude reads/writes directly, scales fine for study-scale data
SR algorithm	SM-2	Anki's foundation — simple, well-understood, sufficient for weeks-to-months timelines
Review style	Interstitial	No dedicated "card block" — cards woven into topic transitions, exercise breaks, warm-ups
Exercise model	File-based	User solves in their editor; two versions per concept (guided + from-scratch)
Living documents	Schedule + progress separate from plan	Plan records the commitment (immutable); schedule adapts; progress tracks reality
Daily notes	Write after every task	Context window safety valve — protects against compaction during long sessions