Core Curriculum Verification Register

Core Curriculum Verification Register

Completion record

All 65 claim-level curriculum pages were reviewed line by line on 2026-09-02. Claim-bearing prose, definitions, and comparison tables were checked against the primary or official references defined on each page. Exercises, design prompts, and acceptance criteria were classified as instructions rather than factual claims. Every Mermaid block was checked for agreement with its surrounding explanation and cited concepts.

Method

  1. Inventory every core page carrying the claim-level fact-check scope.
  2. Separate externally verifiable claims from exercises, local inventory observations, and explicitly labeled design guidance.
  3. Narrow or correct unsupported, universal, or obsolete wording.
  4. Add inline source markers adjacent to material claims and direct reference definitions on the same page.
  5. Recheck changing specifications and official product documentation as of 2026-09-02.
  6. Compare every Mermaid node and edge with the described process; record the semantic-review date below each diagram.

A marker count is a mechanical count of [S1] and [S2] occurrences, not a count of unique claims. “Semantic reviews” must equal “Mermaid blocks.” Agentctl output was used as a secondary challenge pass and source-discovery aid, never as evidence by itself.

Register

Page Line review Source-marker occurrences Mermaid blocks Semantic reviews
AI, Machine Learning, Deep Learning, and Generative AI Verified 38 1 1
Data, Training, Validation, and Inference Verified 56 1 1
Neural-Network Basics Verified 40 1 1
Transformers and Attention Verified 32 1 1
Tokens, Embeddings, and Context Verified 44 1 1
Generation, Decoding, and Hallucinations Verified 38 1 1
Model Families and Multimodality Verified 40 1 1
Frame the Job and Success Criteria Verified 46 1 1
Prompt Anatomy Verified 32 1 1
Context Engineering for Users Verified 46 1 1
Verification, Citations, and Abstention Verified 44 1 1
Model Selection, Cost, and Privacy Verified 38 1 1
Repeatable User Workflows Verified 32 1 1
What Is an AI Harness? Verified 38 2 2
AI Harness Reference Architecture Verified 38 1 1
Instruction and Context Assembly Verified 44 1 1
Model Adapters and Routing Verified 46 1 1
Tools and Structured Outputs Verified 38 1 1
State, Memory, and Checkpoints Verified 36 1 1
Permissions, Sandboxing, and Human Approval Verified 36 1 1
Retries, Idempotency, and Recovery Verified 32 1 1
Observability, Cost, and Run Records Verified 44 1 1
Harness Contract and Lifecycle Verified 42 1 1
Agents Versus Workflows Verified 34 1 1
AI Workflow Patterns Verified 36 1 1
Planning Loops and State Graphs Verified 34 1 1
Tool Use and the Model Context Protocol Verified 34 2 2
Multi-Agent Systems Verified 32 1 1
Human-in-the-Loop Control Verified 36 1 1
Skills and Capability Packages Verified 36 1 1
RAG Reference Pipeline Verified 26 2 2
Ingestion, Chunking, and Metadata Verified 42 1 1
Retrieval, Hybrid Search, and Reranking Verified 32 1 1
Grounding, Citations, and RAG Evaluation Verified 34 1 1
Memory, RAG, CAG, and Fine-Tuning Verified 34 1 1
Graph and Structured Retrieval Verified 32 1 1
API-Hosted Versus Local Models Verified 40 1 1
AI Frameworks Without Lock-In Verified 36 1 1
Fine-Tuning, PEFT, and Distillation Verified 32 1 1
Inference, Quantization, and Serving Verified 32 1 1
Multimodal Application Pipelines Verified 22 1 1
AI Production Release Lifecycle Verified 38 1 1
Evaluation Layers and Metrics Verified 34 1 1
Evaluation Datasets, Rubrics, and Edge Cases Verified 32 1 1
Deterministic Checks, Model Judges, and Humans Verified 30 1 1
Online Monitoring and Drift Verified 32 1 1
Experiments, Versioning, and Regression Gates Verified 34 1 1
AI Incidents and Change Management Verified 34 1 1
Threat-Model the Complete AI System Verified 20 1 1
Prompt Injection and Untrusted Content Verified 38 1 1
Data Privacy and Sensitive Information Verified 36 1 1
Model, Data, and Software Supply Chain Verified 32 1 1
Excessive Agency and Tool Risk Verified 36 1 1
Govern, Map, Measure, and Manage AI Risk Verified 28 1 1
Event-Driven AI Automation Verified 18 1 1
Scheduled AI Research and Reporting Verified 36 1 1
n8n AI Workflow Safety and Operations Verified 38 1 1
Lab 1 — Prompt and Verification Verified 30 1 1
Lab 2 — Local RAG Verified 20 1 1
Lab 3 — Read-Only Tool Agent Verified 36 1 1
Lab 4 — Production Harness Capstone Verified 24 1 1
AI Authoritative Source Ledger Verified 22 1 1
Dumpster AI Research Map Verified 56 1 1
AI Glossary Verified 64 1 1
agentctl Research Audit — Learn AI Curriculum Verified 58 1 1

Totals

The authoritative source list and dated API/framework checks are maintained in the source ledger. Legacy files are tracked separately in the legacy fact-check matrix.