Build expertise. Validate it. Own it.
╭────────────────────────────────────────────╮
│ s u k i │
│ Build expertise. Validate it. Own it. │
╰────────────────────────────────────────────╯
Suki turns an AI coding agent into a learning partner that doesn't trust you. Not in a moral sense. It refuses to believe you know something just because you read it, watched a course, or nodded along with a tutorial.
Learning isn't consuming content. It's being forced to say it back, from your own head, until it holds. Suki makes you do that, then it remembers what you're weak at and brings it back before it fades.
🔒 Local-first · 🛡️ Private by default · 🧩 3 steps, one command · 🐍 Python 3 + your agent
┌──────────────────────────────────────────────┐
│ │
▼ │
┌──────────┐ ┌──────────┐ ┌──────────┐ │
│ curriculum│ ──► │ probe │ ──► │ book │ │
│ build the │ │ test it │ │ publish │──────┘
│ path │ │ & fix │ │ it │
└──────────┘ └──────────┘ └──────────┘
— design the definitive learning path: the chapters, the order, what "mastered" means for your topic./suki curriculum <topic>
— build and verify understanding chapter by chapter. Force a restatement, test it, repair the cracks, then schedule reviews so it doesn't rot./suki probe <topic> [ch]
— render your curriculum, working models, and probe history into a publication-quality teaching book./suki book <topic>
Work compounds: each step writes an artifact to ~/.suki/topics/<slug>/
that the next step reads. Probe again next week, and the book reflects it.
/suki curriculum active-directory
→ tier 1: The basics (what AD is, how a domain works)
→ tier 2: Core mechanics (Kerberos, tickets, trusts)
→ tier 3: Attacks (Kerberoasting, golden tickets, delegation)
→ tier 4: Defenses + detection
→ tier 5: Expert edge (red-team tradecraft, real-world ops)
+ what "mastered" means at every chapter
+ the order that actually builds understanding
Suki designs the path from absolute basics to the deepest expert end of the
field, then writes it as chapter-by-chapter markdown plus a
curriculum.json
that every later step reads.
| Command | What it does |
|---|---|
/suki curriculum <topic> |
|
| Design a definitive learning path: chapters, order, mastery checks | |
/suki learn <draft> |
|
| Review a chapter draft from a serious learner's perspective |
/suki probe active-directory 1.1
"Explain Kerberos authentication in your own words."
→ you state your picture
→ suki probes it for cracks
→ weak spot? a targeted drill fixes it right there
→ mastered? revisit in 3 → 10 → 30 → 90 days
The model has to be in your words, not the book's. If it doesn't hold up, the gap is exposed and repaired inline. What you're weak at is remembered and resurfaced before it fades. Nothing is graded against you; everything is recorded as an artifact that compounds.
| Command | What it does |
|---|---|
/suki probe <topic> [ch] |
|
| Build + verify understanding chapter by chapter; repair cracks; spaced repetition |
/suki book active-directory # -> ~/.suki/topics/active-directory/book/book.pdf
→ your curriculum, as one part per tier
→ your working models, persisted
→ your probe + remediation history
→ typeset with pandoc + LaTeX (KOMA scrbook, Palatino)
Everything you built and validated becomes a real teaching book: the same material other learners can use, and the proof of what you actually hold.
| Command | What it does |
|---|---|
/suki book <topic> |
|
| Render the topic as a publication-quality teaching book |
One command, /suki
, is the status dashboard and the router to everything else. It knows where you left off and what's due next.
╭──────────────────────────────────────────╮
│ s u k i │
│ Build expertise. Validate it. Own it. │
╰──────────────────────────────────────────╯
📊 Status
────────────────────────────────────────────
Learning 3 topics · 1 due for review
⏳ Due for review
────────────────────────────────────────────
• active-directory 2.1 (Kerberos auth) due today
• python 3.1 (Asyncio) due in 2d
Next step
→ /suki probe active-directory 2.1
────────────────────────────────────────────
pip install suki
suki install --all # links the skills into opencode, claude, and codex
That's it. Restart your agent, then:
/suki curriculum python # build the path
/suki probe python 1.1 # build + verify understanding
/suki book python # publish what you mastered
| Skill | Role |
|---|---|
suki |
|
| Status dashboard + router (single entry point) | |
curriculum |
|
| Design a definitive learning path | |
learn |
|
| Review material from a learner's perspective | |
probe |
|
| Build + verify understanding, repair cracks, spaced repetition | |
book |
|
| Render a topic as a teaching book |
Everything lives under ~/.suki/
, one folder per topic. Current state is JSON; history is append-only JSONL. Nothing is ever deleted, so the full arc of your learning is recoverable.
~/.suki/
└── topics/<slug>/ # curriculum.json, mastery.json, probes, book/
The suki
CLI manages the stack:
suki install [--opencode|--claude|--codex|--all]
— link skills into your agentsuki topics
— list topicssuki status [topic]
— progress + spaced-repetition due statussuki book <topic>
— render a topic as a book
pip install suki
suki install --all
That's the whole setup. pip install suki
gives you the suki
command;
suki install
links all five skills (suki
, curriculum
, learn
,
probe
, book
) into your agent. Default to --all
, or pick just your
agent with --opencode
, --claude
, or --codex
.
Restart your agent after installing, then use /suki
as the single entry
point. It routes subcommands: /suki curriculum <topic>
,
/suki learn <draft>
, /suki probe <topic> [ch]
, /suki book <topic>
,
or /suki
alone for the status dashboard.
From source (contributing):
git clone <this repo> && cd suki
pip install -e .
The core skills need nothing beyond the agent and the suki
CLI (Python 3).
/suki book
also needs pandoc
and a LaTeX distribution with xelatex
:
brew install pandoc
brew install --cask mactex-no-gui # or any TeX Live install
Everything stays on your machine. Curriculum, probe history, and the books you publish: none of it leaves your computer. The only external calls are the LLM/harness and web fetches.
Brutal honesty first: Know the truth about your model** The model is in your words**: You restate it; it becomes yours** Learn for keeps**: Spaced repetition over cramming** Everything is an artifact**: Named files, compounded across sessions** Privacy by default**: Your data never leaves your machine
Suki: Build expertise. Validate it. Own it.