Add final rubric10 evaluation prompt (ready to use)

Created concise, actionable rubric10 evaluation prompt with:

✅ Schema compliance requirements
✅ Post-processing workflow (fix_evaluation_scores.py)
✅ Schema validation step
✅ Correct field names (summary_scores, element_scores, sub_elements)
✅ Quality checklist
✅ Step-by-step workflow
✅ Critical field name warnings

Key differences from original prompt:
- OLD: No schema → inconsistent JSON structures
- NEW: Strict schema compliance (rubric10_semantic_schema.json)
- OLD: No post-processing → math errors
- NEW: Mandatory fix_evaluation_scores.py step
- OLD: Variable field names → HTML renderer fails
- NEW: Exact field names specified (summary_scores)

This prompt ensures all 4 evaluations will have:
1. Consistent structure
2. Correct scores
3. All required metadata
4. Compatible with HTML renderer

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

Author: realmarcin

RUBRIC10_EVALUATION_PROMPT_FINAL.md

@@ -0,0 +1,238 @@
+# Rubric10-Semantic Evaluation: Schema-Compliant Workflow
+
+Using the d4d-rubric10-semantic agent, evaluate the 4 claudecode_agent concatenated D4D files with **strict schema compliance** to ensure consistent JSON structure and accurate scoring.
+
+## ⚠️ Critical Requirements
+
+**All evaluations MUST conform to the JSON schema:**
+- **Schema**: `src/download/prompts/rubric10_semantic_schema.json`
+- **Post-processing**: Score corrections via `scripts/fix_evaluation_scores.py` (REQUIRED)
+- **Validation**: All outputs validated against schema before HTML generation
+
+## Scope
+
+**Concatenated D4D files (claudecode_agent method only):**
+- `data/d4d_concatenated/claudecode_agent/AI_READI_d4d.yaml`
+- `data/d4d_concatenated/claudecode_agent/CHORUS_d4d.yaml`
+- `data/d4d_concatenated/claudecode_agent/CM4AI_d4d.yaml`
+- `data/d4d_concatenated/claudecode_agent/VOICE_d4d.yaml`
+
+## For Each File
+
+### 1. Read the D4D YAML File
+```python
+import yaml
+with open(f"data/d4d_concatenated/claudecode_agent/{project}_d4d.yaml") as f:
+    d4d_data = yaml.safe_load(f)
+```
+
+### 2. Apply Rubric10-Semantic Quality Assessment
+
+**Standard**: 10 elements, 50 sub-elements (binary 0/1 scoring, max 50 points)
+
+**Semantic Enhancements**:
+- Correctness validation (DOI/grant/RRID formats)
+- Consistency checking (dates, affiliations, funding)
+- Semantic understanding (content accuracy)
+- URL validity checking
+
+### 3. Generate Schema-Compliant JSON Evaluation
+
+**REQUIRED Top-Level Fields:**
+```json
+{
+  "rubric": "rubric10-semantic",
+  "version": "1.0",
+  "d4d_file": "data/d4d_concatenated/claudecode_agent/{PROJECT}_d4d.yaml",
+  "project": "{PROJECT}",
+  "method": "claudecode_agent",
+  "evaluation_timestamp": "2025-12-23T21:45:00Z",
+  "model": {
+    "name": "claude-sonnet-4-5-20250929",
+    "temperature": 0.0,
+    "evaluation_type": "semantic_llm_judge"
+  },
+  "summary_scores": {
+    "total_score": 0,
+    "total_max_score": 50,
+    "overall_percentage": 0.0,
+    "grade": "A+"
+  },
+  "element_scores": [...],
+  "semantic_analysis": {...},
+  "recommendations": [...]
+}
+```
+
+**Element Scores Structure** (10 elements, each with 5 sub-elements):
+```json
+"element_scores": [
+  {
+    "id": 1,
+    "name": "Dataset Discovery and Identification",
+    "description": "Can a user or system discover and uniquely identify this dataset?",
+    "sub_elements": [
+      {
+        "name": "Persistent Identifier (DOI, RRID, or URI)",
+        "score": 1,
+        "evidence": "doi: https://doi.org/...",
+        "quality_note": "DOI present and properly formatted",
+        "semantic_validation": "DOI format validated"
+      }
+      // ... 4 more sub-elements
+    ],
+    "element_score": 5,
+    "element_max": 5
+  }
+  // ... 9 more elements
+]
+```
+
+**Semantic Analysis Structure:**
+```json
+"semantic_analysis": {
+  "issues_detected": [
+    {
+      "type": "format_error",
+      "severity": "warning",
+      "description": "...",
+      "fields_involved": ["doi"],
+      "recommendation": "..."
+    }
+  ],
+  "consistency_checks": {
+    "passed": ["dates_chronological"],
+    "failed": ["funding_grant_mismatch"],
+    "warnings": ["missing_orcid"]
+  },
+  "correctness_validations": {
+    "doi_format": "valid",
+    "grant_format": "valid",
+    "rrid_format": "missing",
+    "url_validity": "all_valid"
+  },
+  "semantic_insights": "..."
+}
+```
+
+### 4. Save Evaluation JSON
+**Output**: `data/evaluation_llm/rubric10_semantic/concatenated/{PROJECT}_claudecode_agent_evaluation.json`
+
+## After Evaluating All 4 Files
+
+### 5. Run Score Fixing Script (REQUIRED)
+
+```bash
+poetry run python scripts/fix_evaluation_scores.py \
+  --input-dir data/evaluation_llm/rubric10_semantic/concatenated
+```
+
+**What This Does:**
+- Recalculates `summary_scores.total_score` by summing all sub-element scores
+- Fixes `summary_scores.overall_percentage`
+- Corrects any LLM calculation errors
+- Adds any missing required fields
+
+### 6. Validate Against Schema
+
+```bash
+poetry run python scripts/validate_evaluation_schema.py \
+  data/evaluation_llm/rubric10_semantic/concatenated/*_claudecode_agent_evaluation.json \
+  --schema src/download/prompts/rubric10_semantic_schema.json
+```
+
+**All 4 files MUST validate successfully before proceeding.**
+
+### 7. Generate HTML for Each Evaluation
+
+```bash
+poetry run python scripts/render_evaluation_html_rubric10_semantic.py \
+  data/evaluation_llm/rubric10_semantic/concatenated/ \
+  data/d4d_html/concatenated/claudecode_agent/
+```
+
+**Output Files:**
+- `data/d4d_html/concatenated/claudecode_agent/AI_READI_evaluation.html`
+- `data/d4d_html/concatenated/claudecode_agent/CHORUS_evaluation.html`
+- `data/d4d_html/concatenated/claudecode_agent/CM4AI_evaluation.html`
+- `data/d4d_html/concatenated/claudecode_agent/VOICE_evaluation.html`
+
+### 8. Generate Summary Report
+
+**Create**: `data/evaluation_llm/rubric10_semantic/concatenated/summary_report.md`
+
+**Contents:**
+- Comparison table showing all 4 projects
+- Element-level performance breakdown
+- Semantic analysis highlights (issues by type)
+- Common strengths and weaknesses
+- Key insights and recommendations
+
+## Settings
+
+- **Temperature**: 0.0 (fully deterministic)
+- **Model**: claude-sonnet-4-5-20250929
+- **Agent**: d4d-rubric10-semantic
+- **Rubric**: `data/rubric/rubric10.txt` + semantic enhancements
+- **Schema**: `src/download/prompts/rubric10_semantic_schema.json`
+
+## Expected Output Structure
+
+```
+data/evaluation_llm/rubric10_semantic/concatenated/
+├── AI_READI_claudecode_agent_evaluation.json  ✅ Schema-compliant, scores corrected
+├── CHORUS_claudecode_agent_evaluation.json    ✅ Schema-compliant, scores corrected
+├── CM4AI_claudecode_agent_evaluation.json     ✅ Schema-compliant, scores corrected
+├── VOICE_claudecode_agent_evaluation.json     ✅ Schema-compliant, scores corrected
+└── summary_report.md
+
+data/d4d_html/concatenated/claudecode_agent/
+├── AI_READI_evaluation.html     (updated with corrected scores)
+├── CHORUS_evaluation.html       (updated with corrected scores)
+├── CM4AI_evaluation.html        (updated with corrected scores)
+└── VOICE_evaluation.html        (updated with corrected scores)
+```
+
+## Quality Checklist
+
+Before considering evaluation complete, verify:
+
+- [ ] All 4 JSON files conform to `rubric10_semantic_schema.json`
+- [ ] `fix_evaluation_scores.py` completed successfully
+- [ ] All 4 files pass schema validation
+- [ ] HTML files generated without errors
+- [ ] HTML displays correct scores (matching `summary_scores.total_score`)
+- [ ] `summary_report.md` created with all sections
+- [ ] No "null" values in required fields
+- [ ] All sub-element scores are 0 or 1 (binary)
+- [ ] `element_score` equals sum of sub-element scores
+- [ ] `summary_scores.total_score` equals sum of all element scores
+
+## Key Field Names (CRITICAL)
+
+**Use EXACTLY these field names:**
+
+✅ **CORRECT**: `summary_scores` (with `total_score`, `total_max_score`, `overall_percentage`)
+❌ **WRONG**: `overall_scores`, `overall_score`, `overall_summary`, `overall_assessment`
+
+✅ **CORRECT**: `element_scores` (array of 10 elements)
+❌ **WRONG**: `elements`, `element_evaluations`
+
+✅ **CORRECT**: `sub_elements` (array of 5 per element)
+❌ **WRONG**: `sub_element_scores`, `subelements`
+
+**Note**: Using incorrect field names will cause HTML generation to fail.
+
+## References
+
+- **Schema**: `src/download/prompts/rubric10_semantic_schema.json`
+- **Output Example**: `src/download/prompts/rubric10_output_format.json`
+- **Rubric**: `data/rubric/rubric10.txt`
+- **HTML Renderer**: `scripts/render_evaluation_html_rubric10_semantic.py`
+- **Fix Script**: `scripts/fix_evaluation_scores.py`
+- **Issue Report**: `RUBRIC10_ISSUES_REPORT.md` (why regeneration needed)
+- **Full Documentation**: `RUBRIC10_UPDATED_PROMPT.md`
+
+---
+
+Process all 4 files systematically and generate structured, schema-compliant evaluation outputs with semantic analysis for the latest improved D4D datasheets.