Codebase cleanup: schema fixes, CI, and documentation improvements

Schema fixes:
- Add missing allOf conditions for horizontalRule and break block types
- Fix tableCell children to require blocks only (not textNode), resolving oneOf ambiguity
- Fix cross-schema references to use full URLs instead of relative paths

CI/CD:
- Add GitHub Action for automated schema and example validation
- Handle cross-schema references with -r flag for annotations and phantoms schemas

Documentation:
- Create comprehensive example document demonstrating all block types
- Update spec to reflect tableCell block-only children requirement
- Add tableCell content rule and update examples in spec
- Update Editors field in introduction
- Add strategic insights summary to design-decisions.md
- Archive strategy conversation notes to docs/archive/

Author: gvonness-apolitical

.github/workflows/validate-schemas.yml

@@ -0,0 +1,82 @@
+name: Validate Schemas
+
+on:
+  push:
+    paths:
+      - 'schemas/**'
+      - 'examples/**'
+      - '.github/workflows/validate-schemas.yml'
+  pull_request:
+    paths:
+      - 'schemas/**'
+      - 'examples/**'
+      - '.github/workflows/validate-schemas.yml'
+
+jobs:
+  validate:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: actions/setup-node@v4
+        with:
+          node-version: '20'
+
+      - name: Install ajv-cli
+        run: npm install -g ajv-cli ajv-formats
+
+      - name: Validate schemas compile
+        run: |
+          # Compile standalone schemas
+          for f in schemas/anchor.schema.json schemas/asset-index.schema.json \
+                   schemas/content.schema.json schemas/dublin-core.schema.json \
+                   schemas/manifest.schema.json schemas/precise-layout.schema.json \
+                   schemas/presentation.schema.json schemas/provenance.schema.json; do
+            echo "Compiling $f..."
+            ajv compile -s "$f" --spec=draft2020 --strict=false
+          done
+
+          # Compile schemas with cross-references (need -r for referenced schemas)
+          echo "Compiling schemas/annotations.schema.json..."
+          ajv compile -s schemas/annotations.schema.json \
+            -r schemas/anchor.schema.json --spec=draft2020 --strict=false
+          echo "Compiling schemas/phantoms.schema.json..."
+          ajv compile -s schemas/phantoms.schema.json \
+            -r schemas/anchor.schema.json --spec=draft2020 --strict=false
+
+      - name: Validate example documents
+        run: |
+          # Validate simple-document
+          ajv validate -s schemas/manifest.schema.json \
+            -d examples/simple-document/manifest.json \
+            --spec=draft2020 --strict=false
+          ajv validate -s schemas/content.schema.json \
+            -d examples/simple-document/content/document.json \
+            --spec=draft2020 --strict=false
+          ajv validate -s schemas/dublin-core.schema.json \
+            -d examples/simple-document/metadata/dublin-core.json \
+            --spec=draft2020 --strict=false
+
+          # Validate signed-document
+          ajv validate -s schemas/manifest.schema.json \
+            -d examples/signed-document/manifest.json \
+            --spec=draft2020 --strict=false
+          ajv validate -s schemas/content.schema.json \
+            -d examples/signed-document/content/document.json \
+            --spec=draft2020 --strict=false
+          ajv validate -s schemas/dublin-core.schema.json \
+            -d examples/signed-document/metadata/dublin-core.json \
+            --spec=draft2020 --strict=false
+
+          # Validate comprehensive-document (if exists)
+          if [ -d examples/comprehensive-document ]; then
+            ajv validate -s schemas/manifest.schema.json \
+              -d examples/comprehensive-document/manifest.json \
+              --spec=draft2020 --strict=false
+            ajv validate -s schemas/content.schema.json \
+              -d examples/comprehensive-document/content/document.json \
+              --spec=draft2020 --strict=false
+            ajv validate -s schemas/dublin-core.schema.json \
+              -d examples/comprehensive-document/metadata/dublin-core.json \
+              --spec=draft2020 --strict=false
+          fi

…trategy/conversation-notes-2025-01-25.md …archive/conversation-notes-2025-01-25.mddocs/strategy/conversation-notes-2025-01-25.md renamed to docs/archive/conversation-notes-2025-01-25.md docs/strategy/conversation-notes-2025-01-25.md renamed to docs/archive/conversation-notes-2025-01-25.md

@@ -499,3 +499,80 @@ Should the format support DRM?
 
 **Status**: Explicitly out of scope
 **Considerations**: Opens governance issues; encryption provides confidentiality
+
+---
+
+## Strategic Insights
+
+This section captures key strategic insights from early design discussions that inform the specification's direction and adoption approach. Full discussion notes are archived in `docs/archive/`.
+
+### SI-001: Technical Merit vs Adoption
+
+**Insight**: Technical problems are real and documented, but technical merit is approximately 20% of what determines format success. The other 80% is ecosystem, timing, and adoption strategy.
+
+**Evidence**:
+- PDF signature vulnerabilities proven in published security research (21 of 22 desktop viewers vulnerable)
+- View/edit divide creates universal workflow friction
+- Multi-billion dollar AI extraction industry exists because PDF structure is unreliable
+
+**Implication**: Technical case is necessary but not sufficient. A spec without robust tooling is just documentation.
+
+---
+
+### SI-002: Beachhead Strategy — Academia First
+
+**Insight**: Academia is the optimal initial adoption target, with legal/enterprise as a secondary market.
+
+**Why Academia**:
+- Lower switching costs (no enterprise contracts, IT approval chains)
+- Cultural alignment with open standards
+- Acute pain points: citations as flat text, figures as inaccessible blobs, unreliable text extraction
+- LaTeX users prove academics tolerate complexity for better output
+- Natural integration points: Overleaf, Zotero, Pandoc, Jupyter
+- Long-term pipeline: grad student → professor → journal editor → department mandate (10-15 year arc)
+
+**Why Legal Secondary**:
+- High pain point but high friction (entrenched tooling, tech-averse users)
+- Better play: become "killer feature" for tooling vendor entering legal market
+- "If lawyers trust it for contracts" provides powerful social proof
+
+---
+
+### SI-003: Development Philosophy
+
+**Insight**: Start solo, design for OSS. Empty repos don't attract contributors; working code does.
+
+**Approach**:
+1. Begin implementation alone to move fast and establish patterns
+2. Design for OSS from day one (clear architecture, good docs, contribution points)
+3. Open implementation once there's something functional to contribute to
+4. The spec is already open — that's the legitimacy part
+
+**Rationale**: Speed now, community later. Avoid "design by committee" early.
+
+---
+
+### SI-004: Implementation Priorities
+
+**Insight**: Pandoc integration is the highest-leverage early move.
+
+**Build Order**:
+1. **cdx-core** (Rust library) — Foundation everything else builds on
+2. **cdx-cli** — Dogfoods the core library, essential for tooling development
+3. **Pandoc writer** — Markdown → Codex (the academia unlock)
+4. **Web viewer** — cdx-core compiled to WASM (zero-install demonstration)
+
+**Why Pandoc**: Academics don't adopt new editors, they adopt new export targets. A Pandoc writer fits existing workflows with zero friction for authors.
+
+---
+
+### SI-005: Spec Evolution Principles
+
+**Insight**: Specs accumulate ad-hoc solutions. Catching inconsistencies early (pre-v1.0) and unifying is much cheaper than fixing later.
+
+**Lessons Learned**:
+- "Where does this live?" is the critical question for new features (inside vs outside hashing boundary)
+- Contradictions hide in prose — normative algorithms trump aspirational text
+- Each fix pulls on connected threads — changes are individually clean but interdependent
+
+**Example**: Three extensions independently invented sub-block addressing. Unifying into the anchor system prevented fragmentation.