Enhance MetaTraits transform with chemical mapping and expanded METPO predicates

.claude/skills/audit-mappings/SKILL.md

@@ -0,0 +1,129 @@
+---
+name: audit-mappings
+description: Audit transform code for hardcoded ontology mappings and generate coverage report
+---
+
+# Audit Mappings Skill
+
+Scans KG-Microbe transform code to identify hardcoded ontology mappings and generates a comprehensive audit report. This skill helps maintain data-driven architecture by detecting inline CURIE mappings that should be moved to mapping files.
+
+## What This Skill Audits
+
+### Python Code Patterns
+- **Hardcoded dictionaries** with CURIE values (e.g., `{"trait": "METPO:12345"}`)
+- **Inline string assignments** with ontology prefixes (CHEBI:, GO:, EC:, METPO:, etc.)
+- **Mapping dictionaries** embedded in transform code
+- Filters out false positives: imports, comments, docstrings, configuration constants
+
+### Mapping Files
+- **TSV files** in `kg_microbe/transform_utils/*/mappings/*.tsv`
+- **YAML files** like `custom_curies.yaml`
+- **JSON files** matching pattern `*mapping*.json`
+- Counts entries and categorizes by type
+
+## Usage
+
+### Audit all transforms
+```bash
+/audit-mappings
+```
+
+### Audit specific transform
+```bash
+/audit-mappings --transform metatraits
+```
+
+### Generate detailed report with code snippets
+```bash
+/audit-mappings --transform bacdive --verbose
+```
+
+### Generate markdown report
+```bash
+/audit-mappings --format md > mapping_audit_report.md
+```
+
+### Inventory mapping files only (skip code scanning)
+```bash
+/audit-mappings --mapping-files-only
+```
+
+## Options
+
+- `--transform NAME` - Audit specific transform (default: all)
+- `--format {text,json,md}` - Output format (default: text)
+- `--verbose` - Include code snippets and line numbers
+- `--mapping-files-only` - Only scan mapping files, skip Python code
+
+## Output Format
+
+### Text Format (default)
+```
+=== Hardcoded Mapping Audit Report ===
+Date: 2026-04-06
+
+Transform: metatraits
+  Python hardcoded mappings: 1
+    - metatraits.py:52-95 (METPO_TO_BIOLINK_PREDICATE, 44 entries)
+
+  Mapping files: 5
+    - chemical_name_synonyms.tsv (44 entries)
+    - enzyme_name_to_go.tsv (34 entries)
+    - special_chemical_mappings.tsv (35 entries)
+    - ec2go.txt (4,822 entries)
+
+  Status: ✅ 99.97% data-driven
+
+---
+Summary:
+  Total transforms scanned: 20
+  Transforms with hardcoded mappings: 15
+  Total mapping files: 25
+  Total mapping entries: 5,200+
+```
+
+### JSON Format
+```json
+{
+  "report_date": "2026-04-06",
+  "transforms": [
+    {
+      "name": "metatraits",
+      "hardcoded_mappings": [...],
+      "mapping_files": [...],
+      "data_driven_percentage": 99.97
+    }
+  ],
+  "summary": {...}
+}
+```
+
+## Classification Rules
+
+### Data-driven (Good) ✅
+- Mappings loaded from TSV/YAML/JSON files
+- Dynamic lookups via OAK/ChemicalMappingLoader
+- Predicate lookups via resolver methods
+
+### Hardcoded (Flag) ⚠️
+- Inline dictionaries with >5 CURIE mappings
+- String literals with CURIEs in business logic
+- Should be migrated to mapping files
+
+### Acceptable Hardcoded
+- API endpoints and URLs
+- Configuration constants (paths, file names)
+- Schema-level mappings (e.g., METPO → Biolink predicates)
+- Fallback placeholders for ontology gaps
+
+## Use Cases
+
+- **Audit data-driven compliance** before releases
+- **Identify migration candidates** for refactoring
+- **Track mapping coverage** across transforms
+- **Prevent regressions** to hardcoded patterns
+- **CI/CD quality checks** as part of test suite
+
+## Implementation
+
+See `audit_mappings.py` for the scanning logic.