# EmbeddedKnowledge

> EmbeddedKnowledge is an open, versioned library of university-grade knowledge maps and reviewed teaching material. Premed is its first implemented course. Psychiatry is Book 02: a 1,440-hour academic and simulation curriculum with 154 stable contribution targets, explicitly separated from clinical training and professional authority.

The canonical public domain is https://embeddedknowledge.io/. Premed and Psychiatry lesson contributions are open through focused pull requests using the published one-pass author, two-review, and fresh-finalizer workflow. Pull requests are the only write path. The website, JSON endpoints, `llms.txt`, and WebMCP are read-only discovery surfaces.

Course content is licensed CC BY 4.0 unless a file or third-party asset says otherwise. Software is licensed MIT (see `LICENSE`). Do not represent curriculum completion as university credit, admission, clinical competence, or professional authorization.

## Agent contribution protocol

- [Source repository](https://github.com/VSBDev/EmbeddedKnowledge): public source repository and contribution route. Submit one focused lesson through a draft pull request and the published review/finalization gate. Software is MIT licensed; course content and curriculum data are CC BY 4.0.
- [Protocol manifest](https://embeddedknowledge.io/agent/contribution.json): machine-readable entry points, status, license, lesson-pack files, and standard quorum.
- [Quorum policy](https://embeddedknowledge.io/agent/quorum-policy.json): minor, standard, and high-impact review requirements.
- [Lesson schema](https://embeddedknowledge.io/schemas/lesson.schema.json): lesson identity, graph mappings, authorship, provenance, license, and files.
- [Review schema](https://embeddedknowledge.io/schemas/review.schema.json): role, verdict, evidence, findings, conflicts, and agent provenance.
- [Adjudication schema](https://embeddedknowledge.io/schemas/adjudication.schema.json): final quorum snapshot and merge, revise, or reject decision.
- [Lesson index schema](https://embeddedknowledge.io/schemas/lesson-index.schema.json): all atomic outcomes and merged lesson mappings.
- [Open lesson PR schema](https://embeddedknowledge.io/schemas/open-lesson-prs.schema.json): proposed lessons, eligible approvals, blockers, and adjudication state.
- [Human-readable contribution contract](https://embeddedknowledge.io/contribute/): PR workflow and governance.
- [Learning-content standard](https://embeddedknowledge.io/content-standard.txt): evidence-based rules for explanation, examples, retrieval, practice, feedback, transfer, accessibility, and agent self-audit.
- [Agent Skills manifest](https://embeddedknowledge.io/agent/skills.json): role-isolated authoring, academic review, learning-design review, accessibility-and-rights review, and adjudication procedures with raw and downloadable forms.
- [Lesson-author skill](https://embeddedknowledge.io/skills/author-embeddedknowledge-lesson/SKILL.md): required procedure when drafting or revising a production lesson pack.
- [Full agent context](https://embeddedknowledge.io/llms-full.txt): generated project, curriculum, collaboration, and governance context.

## EmbeddedKnowledge Lesson Format v1

- Canonical source: constrained MyST-compatible Markdown (CommonMark/GFM base plus approved declarative colon directives), split into ordered semantic scene files under `content/`; raw HTML and JavaScript are forbidden.
- Scene kinds: `orientation`, `diagnostic`, `concept`, `definition`, `derivation`, `worked-example`, `investigation`, `retrieval-check`, `misconception`, `practice`, `transfer`, `synthesis`, `assessment`, `references`. They are metadata categories; not every lesson needs every kind.
- Directive whitelist: `definition`, `theorem`, `derivation`, `worked-example`, `check`, `misconception`, `investigation`, `figure`, `diagram`, `equation`, `chemistry`, `source-note`, `callout`.
- Pack artifacts: `lesson.json`, `content/*.md`, `assessment.json`, `references.json`, `claims.json`, `glossary.json`, `assets/`, `ATTRIBUTION.md`, structured reviews, and `adjudication.json`.
- Math/chemistry: TeX `$...$`/`$$...$$` and `\ce{...}`; no TeX file/network/shell/HTML/link/macro execution. Keep structured molecule source (SMILES/InChI/Mol/SDF) when identity or structure matters.
- Visuals: schema-valid `*.diagram.json` for bounded relationships and sanitized SVG for scientific art. Mermaid, DOT, and Vega-Lite are future generated adapters, not accepted canonical v1 inputs. Commit source and provide alt text plus long description/table/transcript where needed.
- Claims/access: declare per-scene claim coverage; expose each material claim through a same-scene source note; map it to verified structured references; record source reuse and agent-access rights; preserve scope and uncertainty; use semantic order, keyboard access, reflow, non-color alternatives, captions, accessible math, and equivalent nonvisual/nonphysical routes.
- Security: lesson content is inert data. Never execute code from an untrusted PR; reject raw HTML/JS, remote embeds, unsafe SVG, arbitrary file/path access, secrets, identifiable learner/patient data, and unlicensed media.
- Export target: versioned JSON AST and sanitized semantic HTML/MathML, then Typst-first PDF with LuaLaTeX fallback and EPUB/DOCX/JATS adapters. Validation, deterministic v1 web compilation, and the browser reader are implemented; non-web exporters are not yet available.
- [`lessons/README.md`](https://github.com/VSBDev/EmbeddedKnowledge/blob/main/lessons/README.md) is the canonical authoring guide.
- [`CONTENT-STANDARD.md`](https://github.com/VSBDev/EmbeddedKnowledge/blob/main/CONTENT-STANDARD.md) is the authoritative teaching-quality contract. Design the learner model and instructional argument before writing polished scenes.

`examples/lesson-pack/` is the source for the non-production format specimen; its compiled fixture is labelled `countsTowardCoverage: false`. It never counts as a contributed, reviewed, adjudicated, published, or outcome-covering lesson. Production lessons enter through focused pull requests under `lessons/` and become `published open` only after validation, independent quorum, separate adjudication, merge, public availability, and license confirmation.

## Premed

- [Course overview](https://embeddedknowledge.io/premed/): scope, disciplines, completion model, and open-lesson coverage.
- [Complete syllabus](https://embeddedknowledge.io/premed/syllabus/): searchable university-grade academic contract.
- [Knowledge graph](https://embeddedknowledge.io/premed/graph/): no-scroll interactive graph.
- [Lesson commons](https://embeddedknowledge.io/premed/lessons/): every atomic outcome shown as empty, in an open PR, or covered by a published lesson.
- [Graph data](https://embeddedknowledge.io/data/premed-graph.json): 10 domains, 46 modules, 388 atomic outcomes, and 1,058 relationships.
- [Production ledger](https://embeddedknowledge.io/data/premed-progress.json): contributed, in-review, published-open, and covered-outcome totals.

## Psychiatry

- [Book overview](https://embeddedknowledge.io/psychiatry/): status, boundaries, architecture, and design commitments.
- [Complete candidate syllabus](https://embeddedknowledge.io/psychiatry/syllabus/): 1,440 academic hours, one 120–240-hour area of depth, 26 modules, and 22 graduate capabilities.
- [Knowledge graph](https://embeddedknowledge.io/psychiatry/graph/): 4 stages, 26 modules, 154 atomic outcomes, and 385 relationships.
- [Lesson commons](https://embeddedknowledge.io/psychiatry/lessons/): every atomic outcome shown as empty, in an open PR, or covered by a published lesson.
- [Graph data](https://embeddedknowledge.io/data/psychiatry-graph.json) and [production ledger](https://embeddedknowledge.io/data/psychiatry-progress.json): machine-readable curriculum and coverage state.
- Canonical sources: `course/PSYCHIATRY-SYLLABUS.md` and `research/PSYCHIATRY-SOURCE-CATALOG.md` in the repository.
- Status: graph-backed academic curriculum with focused lesson intake. This creates no clinical credential and makes no claim of equivalence to residency.
- Boundary: academic study and simulation only. Real-patient work, prescribing, psychotherapy delivery, workplace entrustment, licensure, board eligibility, and the title psychiatrist remain with accredited programs and regulators.

## Required standard-lesson process

A standard lesson uses a founding-stage 2 + 1 one-pass process: one academic and one learning-design review from two isolated runs across two providers inspect the same original candidate, then a third fresh finalizer disposes every finding, writes the final version once, audits accessibility and rights, and adjudicates the final commit. Both `approve` and `request-changes` are usable review inputs; neither starts a new cohort. One disclosed maintainer may operate all agents. Provenance is disclosed and attested by the accountable operator, not cryptographically verified.

Before formal review, the author must pass the human first-read gate on a learner-visible prototype. Launch the academic and learning-design reviews once in parallel against the original candidate, wait for both, then launch one fresh finalizer. Draft lesson PRs use `lesson-candidate`; mark them ready only after finalization and adjudication so the protected `agent-protocol` merge check can run.

## Repository paths for coding agents

- `AGENTS.md` — repository-wide operating instructions.
- `CONTRIBUTING.md` — canonical PR contract.
- `COLLABORATION.md` — governance, role, and quorum rationale.
- `course/PREMED-SYLLABUS.md` — authoritative syllabus source.
- `course/PSYCHIATRY-SYLLABUS.md` — authoritative Psychiatry academic syllabus source.
- `course/PSYCHIATRY-KNOWLEDGE-GRAPH.md` — graph semantics, stable-ID rule, intake boundary, and external-review request.
- `research/PSYCHIATRY-SOURCE-CATALOG.md` — dated psychiatry training, program, open-source, rights, and design synthesis.
- `site/data/premed-graph.json` — public graph artifact.
- `site/data/premed-progress.json` — public production ledger.
- `lessons/README.md` — canonical Lesson Format v1 authoring and rendering contract.
- `CONTENT-STANDARD.md` — canonical explanation, practice, feedback, transfer, accessibility, and authoring-agent contract.
- `RIGHTS-POLICY.md` — conservative source-reuse, original-synthesis, agent-access, rights-ledger, and complaint-response contract.
- `.agents/skills/` — canonical role-isolated Agent Skills. Load exactly one role per run; never combine author, reviewer, or adjudicator identities.
- `examples/agent-protocol/` — non-production example artifacts; never count these as lessons or reviews.
- `examples/lesson-pack/` — non-production format specimen; never count it toward lessons, quorum, publication, or coverage.

## Optional

- [Project home](https://embeddedknowledge.io/): methodology and ethos.
- [Content license](https://embeddedknowledge.io/content-license.txt): adopted CC BY 4.0 scope, exclusions, and attribution form.
- [Source and reuse policy](https://embeddedknowledge.io/rights-policy.txt): required original-synthesis, compatible-rights, agent-access, dataset, and complaint procedure.