@thanh01.pmt/curriculum-kit 1.2.0 → 1.2.1

package/dist/ai/index.mjs

@@ -2025,13 +2025,13 @@ Evaluate the candidate artifact across these 6 dimensions:
    - 0 pts: Missing critical sections.
 
 5. **Technical & Conceptual Accuracy (15 pts):**
-   - 15 pts: Code snippets, logic flows, architectural descriptions, and Mermaid diagrams are conceptually sound, valid, and free of syntax hallucinations.
+   - 15 pts: Code snippets, logic flows, architectural descriptions, and Mermaid diagrams are conceptually sound and valid. If \`canonicalExposition\` or \`[KNOWLEDGE_EXPOSITION]\` is provided, content strictly adheres to canonical concepts without contradicting definitions or examples.
    - 8 pts: Minor syntax inaccuracies or unidiomatic code that does not break core concepts.
-   - 0 pts: Severe technical hallucinations, broken logic, or invalid diagrams.
+   - 0 pts: Severe technical hallucinations, broken logic, invalid diagrams, or direct contradiction of canonical exposition knowledge.
 
 6. **Contract Consistency & Zero-Drift (10 pts):**
-   - 10 pts: Satellite artifact strictly adheres to the scope, tech stack, and objectives of Lesson ${lessonId} without introducing unrelated topics.
-   - 0-5 pts: Scope drift or topic divergence detected.
+   - 10 pts: Satellite artifact strictly adheres to the scope, tech stack, and objectives of Lesson ${lessonId} without introducing unrelated topics. If \`canonicalLessonExcerpt\` is present in candidate JSON, all satellite phases/activities strictly operationalize the master Lesson Flow and Activity Sequence without inventing divergent timelines.
+   - 0-5 pts: Scope drift, topic divergence, or disjoint activity sequence detected compared to master lesson.
 
 ## VERDICT RULES
 - **PASS:** Score >= 80 AND Language Adherence = true AND Standards Alignment >= 12 (when STANDARDS GROUND TRUTH is present).