mdmurphy822 · mdmurphy822 · Apr 22, 2026 · Apr 17, 2026 · Apr 17, 2026 · Apr 17, 2026
diff --git a/.gitignore b/.gitignore
@@ -35,8 +35,18 @@ coverage.xml
 # =============================================================================
 # Ed4All - LibV2 Content (preserve structure, exclude data)
 # =============================================================================
+# Course directories under LibV2/courses/ contain user-specific content
+# (chunks.jsonl, graph/, quality/, retrieval/gold_queries.jsonl, ...). These
+# are intentionally not tracked so this repository stays course-agnostic.
+# Reference-retrieval fixtures live in
+# LibV2/tools/libv2/tests/test_eval_harness_retrieval.py; users curate their
+# own per-course gold queries locally (see docs/libv2/reference-retrieval.md
+# "Adding gold queries to your own corpus").
 LibV2/courses/*
 !LibV2/courses/.gitkeep
+# LibV2/catalog/cross_package_concepts.json is regenerated on demand via
+# ``libv2 cross-index`` against whatever courses a user has loaded locally;
+# we do not ship a course-specific snapshot.
 LibV2/catalog/*
 !LibV2/catalog/.gitkeep
 
@@ -102,6 +112,8 @@ training-captures/trainforge/*
 !training-captures/trainforge/.gitkeep
 training-captures/textbook-pipeline/*
 !training-captures/textbook-pipeline/.gitkeep
+training-captures/pipeline/*
+!training-captures/pipeline/.gitkeep
 
 # MCP runtime
 MCP/runtime/
@@ -129,3 +141,10 @@ state/GENERATION_PROGRESS.md
 
 # Examples output
 examples/output/
+
+# Planning artifacts — scratch docs, reviews, sub-plans; kept local-only
+plans/
+
+# Top-level inputs/ — corpus PDFs + local scratch; directory kept, contents ignored
+inputs/*
+!inputs/.gitkeep
diff --git a/CLAUDE.md b/CLAUDE.md
diff --git a/Courseforge/CLAUDE.md b/Courseforge/CLAUDE.md
@@ -86,17 +86,14 @@ USER REQUEST →
 ### Course Creation Agents
 | Agent | Purpose | When to Use |
 |-------|---------|-------------|
-| `exam-research` | Certification objective analysis | Analyzing exam/certification objectives |
 | `requirements-collector` | Course specification gathering | New course projects |
-| `course-outliner` | Course structure and learning objectives | Creating course framework |
+| `course-outliner` | Synthesize canonical `TO-NN` / `CO-NN` objectives from textbook structure; persist `synthesized_objectives.json`. Wave 24+ routes this agent to `plan_course_structure` (no longer just creates project dirs). | Creating course framework |
 | `content-generator` | Educational content creation | Content development (1 file per agent) |
-| `educational-standards` | Pedagogical framework compliance | UDL, ADDIE, Bloom's alignment |
 | `quality-assurance` | Pattern prevention and validation | Quality gates |
 | `oscqr-course-evaluator` | Educational quality assessment | OSCQR evaluation |
 | `brightspace-packager` | IMSCC package generation | Final deployment |
-| `html-design-research` | HTML/CSS design pattern research | Template and design validation |
-| `objective-synthesizer` | Learning objective synthesis | Combining objectives from textbooks |
 | `textbook-ingestor` | Textbook content processing | Entry point for textbook materials |
+| `source-router` | Bind DART source blocks to Courseforge module pages | Source attribution for pipeline runs |
 
 ### Intake & Remediation Agents (NEW)
 | Agent | Purpose | When to Use |
@@ -151,10 +148,9 @@ Task(content-generator, "Create all Week 1 content")  # NEVER DO THIS
 ├── schemas/                     # IMSCC and content schemas
 ├── imscc-standards/             # Brightspace/IMSCC technical specs
 ├── scripts/                     # Automation scripts
-│   ├── imscc-extractor/         # Universal IMSCC extraction (NEW)
-│   ├── dart-batch-processor/    # Automated DART conversion (NEW)
-│   ├── component-applier/       # Interactive component application (NEW)
-│   └── remediation-validator/   # Final quality validation (NEW)
+│   ├── imscc-extractor/         # Universal IMSCC extraction
+│   ├── component-applier/       # Interactive component application
+│   └── remediation-validator/   # Final quality validation
 ├── exports/                     # Generated course packages
 │   └── YYYYMMDD_HHMMSS_name/    # Timestamped project folders
 └── runtime/                     # Agent workspaces (auto-created)
@@ -243,25 +239,75 @@ Courseforge HTML pages embed machine-readable instructional design metadata for
 
 | Attribute | Element | Purpose |
 |-----------|---------|---------|
-| `data-cf-objective-id` | `<li>` (objectives) | Learning objective identifier |
+| `data-cf-role` | `<body>` (template chrome) | Page role classification (e.g. `template-chrome`) |
+| `data-cf-objective-id` | `<li>` (objectives) | Learning objective identifier (canonical `TO-NN` / `CO-NN` pattern) |
 | `data-cf-bloom-level` | `<li>`, `.self-check`, `.activity-card` | Bloom's taxonomy level |
 | `data-cf-bloom-verb` | `<li>` (objectives) | Detected Bloom's verb |
+| `data-cf-bloom-range` | `<section>`, `<h2>` | Section-level Bloom level span (emit-only) |
 | `data-cf-cognitive-domain` | `<li>` (objectives) | Knowledge domain (factual/conceptual/procedural/metacognitive) |
 | `data-cf-content-type` | `<h2>`, `<h3>`, `.callout` | Section content classification |
+| `data-cf-teaching-role` | `<section>`, component wrappers | Pedagogical teaching role |
 | `data-cf-key-terms` | `<h2>`, `<h3>` | Comma-separated term slugs |
+| `data-cf-term` | key-term `<span>` | Individual term slug (emit-only) |
 | `data-cf-component` | `.flip-card`, `.self-check`, `.activity-card` | Interactive component type |
 | `data-cf-purpose` | `.flip-card`, `.self-check`, `.activity-card` | Pedagogical purpose |
 | `data-cf-objective-ref` | `.self-check`, `.activity-card` | Associated learning objective |
+| `data-cf-source-ids` | `<section>`, headings, component wrappers | DART `sourceId`(s) that ground this block. Shape: `dart:{slug}#{block_id}`. Carried through from DART's `data-dart-block-id` when source material is present; elided when no source grounding exists. |
+| `data-cf-source-primary` | `<section>`, headings, component wrappers | The primary `sourceId` for the block (subset of `data-cf-source-ids`) when one source dominates. |
+
+Attributes stop at the **section / component wrapper level** — never on every `<p>` / `<li>` / `<tr>` in prose.
+
+### Wave 35: ancestor-walkable grounding
+
+`ContentGroundingValidator` walks each non-trivial `<p>` / `<li>` /
+`<figcaption>` / `<blockquote>`'s ancestor chain to find the first
+`data-cf-source-ids` attribute. Three emit-side contracts keep that
+walk passing:
+
+1. **Content sections are wrapped in `<section data-cf-source-ids="…">`.**
+   `Courseforge/scripts/generate_course.py::_render_content_sections`
+   now wraps each h2/h3 + paragraph group in a `<section>` wrapper
+   carrying the section's resolved source-ids. Pre-Wave-35 the
+   attribute lived only on the `<h2>` (a DOM sibling of the `<p>`),
+   which the validator's ancestor walk couldn't reach.
+2. **`content_NN` pages inherit `content_01` grounding.**
+   `_page_refs_for` falls back from `content_NN` → `content_01` in
+   the `source_module_map`. The source-router only emits a single
+   per-week `content_01` entry; every generated content page in that
+   week shares the same DART source region, so the fallback is the
+   correct grounding (not a workaround).
+3. **Objectives `<section>` mirrors page-level source-ids.**
+   `ensure_objectives_on_page` scans the page body for the first
+   `<section data-cf-source-ids="…">` wrapper and stamps the same
+   ids onto the injected objectives section. Long synthesized LO
+   statements that exceed the 30-word non-trivial floor otherwise
+   flagged as ungrounded under the ancestor walk.
+
+DART-side slug contract (see `DART/CLAUDE.md`): the `dart:{slug}#{block_id}`
+slug uses `lowercase + space-to-hyphen` normalization (not
+`canonical_slug`'s underscore collapse), matching the validator's
+`_resolve_valid_block_ids` rule.
 
 ### JSON-LD Structured Metadata
 
 Each page includes a `<script type="application/ld+json">` block in `<head>` with:
-- `learningObjectives`: ID, statement, Bloom's level/verb, cognitive domain, assessment suggestions
-- `sections`: Heading, content type, Bloom's range, key terms with definitions
+- `learningObjectives`: ID (canonical `TO-NN` / `CO-NN`), statement, Bloom's level/verb, cognitive domain, assessment suggestions
+- `sections`: Heading, content type, Bloom's range, key terms with definitions, optional per-section `sourceReferences`
 - `misconceptions`: Common misconceptions with corrections
 - `suggestedAssessmentTypes`: Recommended question formats
+- `prerequisitePages`: Cross-page prerequisite refs
+- `sourceReferences`: Optional page-level DART source references (canonical `{sourceId, role, weight?, confidence?, pages?, extractor?}` shape). Page-level JSON-LD `role` is authoritative (`primary` / `contributing` / `corroborating`) and takes precedence over attribute-level roles.
 
-Context namespace: `https://ed4all.dev/ns/courseforge/v1`
+Canonical shape: `schemas/knowledge/courseforge_jsonld_v1.schema.json`. Context namespace: `https://ed4all.dev/ns/courseforge/v1`.
+
+### Learning Objective IDs
+
+Emitted LO IDs follow the pattern `^[A-Z]{2,}-\d{2,}$` from the canonical helper `lib/ontology/learning_objectives.py::mint_lo_id`:
+
+- `TO-NN` — terminal (course-wide) objective.
+- `CO-NN` — chapter-level objective.
+
+Synthesized objectives are persisted to `{project}/01_learning_objectives/synthesized_objectives.json` by the `plan_course_structure` phase in the `textbook_to_course` pipeline. Downstream Trainforge consumers match case-insensitively; the `TRAINFORGE_PRESERVE_LO_CASE` flag preserves the emit case.
 
 ---
 
@@ -341,14 +387,13 @@ Courseforge can import and remediate IMSCC packages from:
 ### Scripts for Course Generation
 | Script | Location | Purpose |
 |--------|----------|---------|
-| `generate_course.py` | `scripts/` | Multi-file weekly course generation with metadata enrichment |
-| `package_multifile_imscc.py` | `scripts/` | Package multi-file course output into IMSCC |
+| `generate_course.py` | `scripts/` | Multi-file weekly course generation. Emits page-level JSON-LD, `course_metadata.json`, prerequisite-page refs, `data-cf-teaching-role`, and `data-cf-source-ids` / page-level `sourceReferences` when DART source material is staged. |
+| `package_multifile_imscc.py` | `scripts/` | Packages multi-file output into IMSCC. Structural validation is on by default (per-week `learningObjectives` must resolve to the week's LO manifest). Auto-discovers `course.json` and bundles `course_metadata.json` at the zip root. Manifest uses IMS Common Cartridge v1.3 namespaces; resources are nested under per-week `<item>` wrappers in the organization tree. **This is the runtime target of the MCP `package_imscc` tool** — `MCP/tools/pipeline_tools.py::_package_imscc` imports and delegates here instead of hand-rolling a ZIP. |
 
 ### Scripts for Intake
 | Script | Location | Purpose |
 |--------|----------|---------|
 | `imscc_extractor.py` | `scripts/imscc-extractor/` | Universal IMSCC parsing |
-| `dart_batch_processor.py` | `scripts/dart-batch-processor/` | Parallel DART conversion |
 | `component_applier.py` | `scripts/component-applier/` | Interactive component application |
 | `remediation_validator.py` | `scripts/remediation-validator/` | Final quality validation |
 

diff --git a/Courseforge/README.md b/Courseforge/README.md
@@ -1,201 +1,29 @@
 # Courseforge
 
 [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
-[![Python 3.9+](https://img.shields.io/badge/python-3.9+-blue.svg)](https://www.python.org/downloads/)
 [![WCAG 2.2 AA](https://img.shields.io/badge/WCAG-2.2%20AA-green.svg)](https://www.w3.org/WAI/WCAG22/quickref/)
 
-**Education Reimagined. Access for All.**
+**Generate LMS-ready course packages from learning objectives or accessible textbook content.**
 
-AI-powered course generation and remediation system that creates and improves accessible, LMS-ready IMSCC packages.
+Courseforge turns a set of objectives (and optional DART-processed textbook HTML) into a complete, accessible, IMSCC-packaged online course. Output includes weekly modules with content pages, activities, self-checks, summaries, and discussions — all WCAG 2.2 AA compliant and ready to import into Brightspace, Canvas, Blackboard, or Moodle. Every page carries rich machine-readable metadata (Bloom's-aligned learning objectives, content types, key terms, misconceptions) so downstream tools can ground assessments and retrieval in cited source material. Courseforge can also ingest an existing IMSCC package from any supported LMS and remediate it to 100% WCAG 2.2 AA compliance.
 
-## Overview
-
-Courseforge uses a multi-agent orchestration system to create high-quality online courses compatible with Brightspace/D2L and other learning management systems. It also provides comprehensive intake and remediation capabilities for existing courses.
-
-### Key Features
-
-**Course Creation**
-- **Simplified Pipeline**: Exam objectives + textbooks → IMSCC package
-- **DART Integration**: Accessibility-first textbook processing
-- **Multi-Agent System**: Specialized agents for each phase of development
-- **WCAG 2.2 AA**: Built-in accessibility compliance
-- **OSCQR Standards**: Educational quality validation
-- **Pattern Prevention**: 22+ error patterns identified and prevented
-
-**Course Intake & Remediation (NEW)**
-- **Universal IMSCC Import**: Canvas, Blackboard, Moodle, Brightspace, Sakai
-- **Automated DART Conversion**: PDFs and Office docs → accessible HTML
-- **AI-Powered Accessibility Fixes**: Alt text, heading structure, contrast
-- **Intelligent Component Styling**: AI-selected interactive elements
-- **100% WCAG 2.2 AA Compliance**: Guaranteed accessibility output
-
-## Quick Start
-
-### Mode 1: Course Creation
-1. Place exam objectives in `inputs/exam-objectives/`
-2. (Optional) Process textbooks through DART and place in `inputs/textbooks/`
-3. Invoke: `exam-research → course-outliner → content-generator → brightspace-packager`
-4. Output: IMSCC file in `exports/YYYYMMDD_HHMMSS_coursename/`
-
-### Mode 2: Course Remediation (NEW)
-1. Place existing IMSCC in `inputs/existing-packages/`
-2. Invoke: `imscc-intake-parser → content-analyzer → remediation agents → brightspace-packager`
-3. Output: Improved IMSCC with 100% accessibility compliance
-
-## Project Structure
-
-```
-Courseforge/
-├── CLAUDE.md              # Orchestration instructions
-├── README.md              # This file
-├── docs/                  # Documentation
-│   ├── troubleshooting.md
-│   ├── workflow-reference.md
-│   └── getting-started.md
-├── agents/                # Agent specifications
-├── inputs/                # Input files
-│   ├── exam-objectives/
-│   ├── textbooks/
-│   └── existing-packages/ # IMSCC for intake (NEW)
-├── scripts/               # Automation scripts (NEW)
-│   ├── imscc-extractor/
-│   ├── dart-batch-processor/
-│   ├── component-applier/
-│   └── remediation-validator/
-├── templates/             # HTML templates
-├── schemas/               # IMSCC schemas
-├── imscc-standards/       # Technical specs
-├── exports/               # Generated packages
-└── runtime/               # Agent workspaces
-```
-
-## Available Agents
-
-### Course Creation
-| Agent | Purpose |
-|-------|---------|
-| `exam-research` | Certification objective analysis |
-| `requirements-collector` | Course specification gathering |
-| `course-outliner` | Structure and learning objectives |
-| `content-generator` | Educational content creation |
-| `educational-standards` | Pedagogical compliance |
-| `quality-assurance` | Pattern prevention |
-| `oscqr-course-evaluator` | Quality assessment |
-| `brightspace-packager` | IMSCC packaging |
-
-### Intake & Remediation (NEW)
-| Agent | Purpose |
-|-------|---------|
-| `imscc-intake-parser` | Universal IMSCC import |
-| `content-analyzer` | Accessibility/quality gap detection |
-| `dart-automation-coordinator` | Automated document conversion |
-| `accessibility-remediation` | WCAG 2.2 AA fixes |
-| `content-quality-remediation` | Educational depth enhancement |
-| `intelligent-design-mapper` | AI component selection |
-
-## Workflow
-
-```
-USER REQUEST →
-  exam-research (analyze objectives) →
-  course-outliner (create structure) →
-  content-generator (create content, 10 agents/batch) →
-  quality-assurance + oscqr-course-evaluator (validate) →
-  brightspace-packager (package) →
-  IMSCC OUTPUT
-```
-
-## Textbook Processing
-
-Textbooks must be processed through DART before use:
+## Quick example
 
 ```bash
-# Set DART_PATH to your DART installation directory
-cd $DART_PATH
-python convert.py textbook.pdf -o /path/to/courseforge/inputs/textbooks/
+# From the repo root, as part of the full Ed4All pipeline:
+ed4all run textbook-to-course --corpus my_textbook.pdf --course-name MY_COURSE_101
 ```
 
-DART produces WCAG 2.2 AA accessible HTML with:
-- Semantic structure
-- Alt text for images
-- MathML for equations
-- Proper heading hierarchy
-
-## Quality Standards
-
-### OSCQR Thresholds
-- Pre-development: 70%
-- Pre-production: 90%
-- Accessibility: 100%
-
-### Pattern Prevention
-Critical patterns addressed:
-- Schema/namespace consistency
-- Assessment XML format (QTI 1.2)
-- Content completeness
-- Organization hierarchy
-
-See `docs/troubleshooting.md` for complete pattern list.
-
-## Documentation
-
-| Document | Purpose |
-|----------|---------|
-| `CLAUDE.md` | Main orchestration instructions |
-| `docs/troubleshooting.md` | Error patterns and solutions |
-| `docs/workflow-reference.md` | Execution protocols |
-| `docs/getting-started.md` | Quick start guide |
-| `agents/*.md` | Individual agent specs |
-
-## Technical Requirements
-
-- Python 3.9+
-- Claude Code access
-- Brightspace/D2L for import testing
-
-## DART Setup (Optional)
-
-DART (Digital Accessibility Remediation Tool) is a separate tool used for converting PDFs and Office documents to accessible HTML. It's **optional** but recommended for textbook processing.
-
-### Installation
-
-DART is available as a separate project. Once installed, configure the environment variable:
-
-```bash
-# Add to your shell profile (.bashrc, .zshrc, etc.)
-export DART_PATH=/path/to/your/DART/installation
-```
-
-### Usage with Courseforge
-
-Once `DART_PATH` is set, Courseforge scripts will automatically detect and use DART for:
-- PDF textbook conversion
-- Office document (Word, PowerPoint) conversion
-- Batch document processing via `scripts/dart-batch-processor/`
-
-### Without DART
-
-Courseforge works without DART for:
-- Course creation from exam objectives
-- IMSCC package generation
-- Course remediation (accessibility fixes to existing HTML)
-
-Only textbook PDF/Office conversion requires DART.
+Finished packages land under `Courseforge/exports/YYYYMMDD_HHMMSS_coursename/`. For intake/remediation of an existing package, drop the IMSCC into `Courseforge/inputs/existing-packages/` and run the `intake_remediation` workflow.
 
-## Compatibility
+## LMS compatibility
 
-- Brightspace/D2L
-- Canvas
-- Blackboard
-- Moodle
-- Any LMS supporting IMSCC 1.1+
+Brightspace / D2L, Canvas, Blackboard, Moodle, Sakai, and any LMS that supports IMSCC 1.1 or later.
 
----
+## More
 
-**Version**: 1.0.0
-**License**: MIT
-**Last Updated**: December 2025
+See [`Courseforge/CLAUDE.md`](CLAUDE.md) for the full agent pipeline, metadata contract (`data-cf-*` + JSON-LD), template components, and quality standards.
 
----
+## License
 
-*Love and be loved, live and be free.*
+MIT