Docs · getting started

From zero to your first lesson

Install it, understand the journey, learn the handful of commands, and skim the honest FAQ. For the full step-by-step install (including Cursor), see INSTALL.md in the repo.

Install

Three ways to run it

When in doubt: the plugin for Claude Code, the skills repo for Cursor or for hacking on the internals. dronAIcharya never installs itself from a web page — you run the command.

1 · As a Claude Code plugin recommended

Everything included: skills, teaching subagents, the anti-drift guardrail hooks, and the optional MCP backend.

in a Claude Code session 
/plugin marketplace add dibakshya01/dronAIcharya
/plugin install dronaicharya@dronaicharya
/dronaicharya:start

Or from your shell: claude plugin marketplace add dibakshya01/dronAIcharya then claude plugin install dronaicharya@dronaicharya --scope user. Update with claude plugin update dronaicharya@dronaicharya; uninstall with claude plugin uninstall dronaicharya@dronaicharya (your .dronacharya/ journey files stay in your projects). Requires a Claude Code recent enough to support plugins — check with claude --version.

2 · As a skills repo transparent / hackable

The easiest route points Claude Code straight at the folder, hooks included:

in your shell 
git clone https://github.com/dibakshya01/dronAIcharya
claude --plugin-dir ./dronAIcharya
# then run /dronaicharya:start (or /start, depending on your version)

Prefer the skills available everywhere? Run bash dronAIcharya/scripts/sync-to-local.sh to copy skills/ into ~/.claude/skills/ and agents/ into ~/.claude/agents/. Note: a bare copy like this does not install the guardrail hooks, so teaching-mode persistence is best-effort — the start / continue skills still carry the teaching contract in-content. Use --plugin-dir or the plugin install to get hooks.

3 · In Cursor

Cursor reads Claude-style skills, agents and AGENTS.md:

in your shell 
git clone https://github.com/dibakshya01/dronAIcharya
bash dronAIcharya/scripts/sync-to-local.sh --cursor
# then open the project in Cursor and ask it to run the start skill

Honest caveat: Claude Code's guardrail hooks do not run in Cursor (a different hook system), so anti-drift / teaching-mode persistence is absent or best-effort there. You still get the full tutor and all four tracks. MCP wiring also differs — follow Cursor's MCP docs for the optional purple-sparrow backend.

The journey

What happens, start to finish

One linear, guardrailed flow. Escape hatches are honored at every point — "pause", "skip ahead", "just fix this quickly", "ELI5", "go deeper", "quiz me", "where am I?".

Greet & start

Paste the repo link, pick an install path, run /dronaicharya:start. A short intro explains what this is and what will happen.

Persona & cadence

Choose a teacher archetype, then how often it stops to teach and check — Thorough, Balanced, or Fast.

Setup check

API key? Editor web / MCP enabled? The tokens and setup skills help if needed, honest about the parts outside the tool's control.

Track & level

Pick Vibecoding, Python, CS, or Cyber & AI Security, and your level: never-coded, some, or confident.

Idea & plan

Bring your own idea or pick a simple-first starter, then co-author a combined BUILD + LEARN plan you approve.

Execute, step by step

Teach → MCQ check → decide the tradeoffs together → build a layer (tracer-bullet first) → explain under the hood → check → advance.

Production path

Architecture, backend (purple-sparrow optional), payments, GitHub, deployment prep — up to your own clicks.

Finish

An "under the hood" recap, a formative mastery review, and what to learn next. Resume anytime with /dronaicharya:continue.

Commands

The skills you can pull any time

As a plugin, each is namespaced /dronaicharya:<name>. The linear journey is driven by one orchestrator; you can also invoke any of these on demand.

Flow
  • start — begin a journey
  • plan — co-author the BUILD + LEARN plan
  • continue — resume where you left off
  • status — progress, pause, disable
Teaching
  • teach — explain the current concept
  • eli5 — explain it very simply
  • deep-dive — go further
  • quiz — check understanding
  • under-the-hood — how it really works
Build / SDLC
  • spec — spec-driven development
  • architect — system design & tradeoffs
  • build-step — build a layer
  • backend — the optional real backend
  • payments · deploy · ship-github
Enablement
  • tokens — get & safely paste an AI API token (you paste it into your editor, never into the chat)
  • setup — enable web / MCP; walk the optional purple-sparrow backend
State

Per-project journey lives in ./.dronacharya/ (plan, progress, journal — gitignored, no secrets). Global mastery lives in ${CLAUDE_PLUGIN_DATA}/profile.json so it carries across projects.

FAQ

Honest answers

Do I need to know how to code already?
No. There's a track and a level for absolute beginners, and it hand-holds the first steps — getting an API token, enabling web / MCP, basic GitHub — broken down as simply as you need, with ELI5 on demand. See the plain-language page.
Will it just write all the code for me?
That's explicitly not the goal. It strongly steers you to understand what's being built, requiring a decision from you at each decision of consequence. But it always obeys "pause / skip / just do it" — you're in control.
Does it send my data or API key anywhere?
No telemetry. Your API key stays on your machine — it never sees or transmits it — and it makes no network calls except the web/doc fetches a lesson explicitly performs. Fetched content is treated as data, not instructions.
What's different in Cursor?
You get the full tutor, all four tracks, and the skills and agents. But Claude Code's guardrail hooks may not run in Cursor, so teaching-mode persistence and the anti-drift nudge are best-effort or absent — a materially lighter, more drift-prone experience.
Is the CS track a full computer-science degree?
No. It's undergraduate foundations and literacy — bits and memory, data structures, Big-O, recursion, OS / networking / database basics, how the web works. No formal proofs, automata, or discrete-math depth.
Do the quizzes certify me?
No. Mastery checks are formative self-checks that help the learning stick and gate advancing. They are not a certificate.
Do I have to set up the purple-sparrow backend?
No — it's entirely optional and advanced. dronAIcharya works fully without it, falling back to conceptual teaching or a simple local store. If you want a real backend to learn on, the setup skill walks you through it. Note purple-sparrow is alpha and shouldn't hold real or regulated data yet.
How do I pause or uninstall?
Run /dronaicharya:status to pause teaching (all hooks go quiet) or see how to disable the plugin. Uninstall with claude plugin uninstall dronaicharya@dronaicharya; your .dronacharya/ journey files stay in your projects.

Ready to take aim?

Install, pick a teacher and a track, and bring an idea.

Install dronAIcharya See how it's built