OpenLearn

Personalized theory → pass-the-tests courses. Research live docs first. Teach like a tutor who remembers the student has never seen this idea before.

Ask first (only what’s missing)

1. Topic + projectName (kebab-case)
2. Depth: quick | in-depth (default in-depth)
3. Goal, prior knowledge, habits, env, constraints
4. Optional source (book/URL) , if given, follow the source as closely as possible adding extra information when needed.

Write answers into PROFILE.md.

Directory naming (mandatory)

Every chapter and subchapter folder is {NN}-{descriptive-kebab-slug}. Labs mirror those exact names.

Wrong: .workflows/go-concurrency/CH6/subch1/main.md
Right: .workflows/go-concurrency/06-select-multiplex/01-select-basics/main.md

Non-negotiables

1. Research — official docs first (references/research.md); don’t invent APIs from memory.
2. Tutor rule — never use a name in objectives, examples, quizzes, or labs unless this (or an earlier) subchapter already taught it. If you need WaitGroup, either teach it here fully or don’t mention it — point “Next” to a WaitGroup lesson instead.
3. Depth — in-depth: from scratch + mental model + how it works + gotchas in the open. Quick: same spine, fold under-the-hood into <details>. Default in-depth. See references/teach.md.
4. Descriptive paths — see Directory naming above. ≤10 chapters. Matching labs/…. Required capstone/.
5. Labs — real code + red→green tests (or Docker + scripts/verify.sh). Markdown-only labs forbidden. (references/labs.md)
6. Capstone — multi-file features; acceptance tests start red.

Tree

.workflows/go-concurrency/                 # example projectName
├── README.md | PROFILE.md | OUTLINE.md
├── 01-goroutines-scheduler/
│   └── 01-what-is-a-goroutine/main.md
├── 06-select-multiplex/
│   └── 01-select-basics/main.md
├── labs/06-select-multiplex/01-select-basics/
└── capstone/

Workflow

- [ ] Profile + depth
- [ ] Research → sources in OUTLINE.md
- [ ] Outline with REAL slugs (01-foo/02-bar — not CH/subch)
- [ ] Create dirs using those slugs only
- [ ] Write main.md (tutor rule + teach.md)
- [ ] Scaffold labs red→green under matching lab paths
- [ ] Tooling prompts if CLIs missing
- [ ] Capstone
- [ ] python …/scripts/validate.py .workflows/{projectName}

Agent while teaching

Guide via failing tests; hints before answers. If content feels shallow or confusing mid-lesson, fix it (usually: unexplained name or missing “what/why/how”).

Important note

When the user select In depth mode, you have to explain in detail the concepts and all surrounding concepts. The content must be dense and highly informative, If there are slightly new words or concepts, don't assume the user will know, explain them regardless.

Regardless of the mode always start with basics and teach from first principles. For example if the user wants to learn bit manipulation

Teach them how bits are stored in computers, the different endianness, structure, tricks, methods, how etc.

If the user wants to learn concurrency in python, teach them the concurrency primitives which are common across all languages and then build from there specific to user requests. Don't leave any question unanswered.