Diagnostic Assessments vs Full Mock Exams in the Claude AI Certification System

In the rohitg00/ai-engineering-from-scratch repository, diagnostic assessments are lightweight, domain-specific checks identified by "kind": "diagnostic", while full mock exams are comprehensive curriculum simulations marked with "kind": "mock".

The Claude AI certification system provides structured evaluation paths to help learners verify their readiness before attempting official certification. Understanding the distinction between diagnostic assessments and full mock exams allows candidates to optimize their study schedules and focus remediation efforts precisely where needed. Both assessment types are defined by JSON configurations stored in the certifications/claude/assessments/ directory.

Core Architectural Differences

The certification framework organizes assessments into two distinct categories based on scope and pedagogical intent.

Purpose and Scope

Diagnostic assessments provide quick, targeted evaluations of specific domains such as model selection, tool use, or governance. These assessments cover only a single domain or small set of related concepts, generating immediate feedback on strengths and weaknesses to guide further study.

Full mock exams deliver comprehensive, end-to-end simulations of the actual certification test. They mirror the structure, timing, and question distribution of the real exam, encompassing all domains defined for the track including multiple-choice, scenario-based, and practical questions.

JSON Structure and Metadata

The system distinguishes assessment types through the "kind" field in JSON configuration files.

Diagnostic files contain:

{
  "kind": "diagnostic",
  "domain": "model_selection",
  "questions": [...]
}

Mock exam files contain:

{
  "kind": "mock",
  "duration_minutes": 120,
  "questions": [...]
}

When to Use Each Assessment Type

Strategic timing maximizes the effectiveness of both diagnostic assessments and full mock exams in your certification journey.

Use diagnostic assessments early in a track, after completing individual lessons or lesson series. These lightweight checks help focus further study on specific areas requiring improvement rather than reviewing material you have already mastered.

Deploy full mock exams only after completing the majority of the track's lessons and diagnostics. These comprehensive simulations serve as final rehearsal tools immediately before attempting the official certification.

Implementation in the Source Code

The rohitg00/ai-engineering-from-scratch repository implements this dual-assessment system through structured JSON files and Python utilities.

File Structure and Naming Conventions

Assessment configurations reside in track-specific subdirectories:

• Diagnostic assessment definition: certifications/claude/assessments/ccar-f/diagnostic.json
• Example full mock exam: certifications/claude/assessments/ccar-f/mock-01.json
• Certification track manifest: certifications/claude/tracks/ccar-f.json

The track manifest lists both assessment types, enabling the system to present appropriate options based on learner progress.

Python Implementation Example

The following Python snippet demonstrates how to load assessment files and distinguish types programmatically:

import json
from pathlib import Path

def load_assessment(track: str, filename: str) -> dict:
    """Load a Claude assessment JSON file."""
    path = Path(__file__).parent / "certifications" / "claude" / "assessments" / track / filename
    with path.open() as f:
        return json.load(f)

def assessment_type(data: dict) -> str:
    """Return a human-readable type based on the 'kind' field."""
    return "Diagnostic" if data.get("kind") == "diagnostic" else "Full Mock Exam"

# Example usage

diagnostic = load_assessment("ccar-f", "diagnostic.json")
mock_exam   = load_assessment("ccar-f", "mock-01.json")

print(f"Diagnostic assessment:  {assessment_type(diagnostic)}")
print(f"Mock exam assessment:   {assessment_type(mock_exam)}")

Summary

• Diagnostic assessments use "kind": "diagnostic" and target single domains with limited question sets for quick progress checks.
• Full mock exams use "kind": "mock" and simulate complete certification tests covering all curriculum domains.
• Both assessment types are stored in certifications/claude/assessments/ with track-specific subdirectories like ccar-f/.
• The load_assessment() function parses JSON configurations while assessment_type() differentiates them via the kind field.
• Diagnostics are best used early and frequently; mock exams should be reserved for final preparation stages.

Frequently Asked Questions

How does the system differentiate between diagnostic and mock exam files?

The system examines the "kind" field in the JSON payload. Files containing "kind": "diagnostic" are processed as lightweight domain checks, while those with "kind": "mock" are treated as full-length practice examinations covering the entire certification curriculum.

Can I take multiple diagnostic assessments before attempting a mock exam?

Yes, the certification framework is designed for iterative learning. You should complete multiple diagnostic assessments across different domains—such as model selection, tool use, and governance—before attempting a full mock exam to ensure comprehensive knowledge coverage.

What is the file path structure for assessment configurations?

Assessment configurations follow the pattern certifications/claude/assessments/{track}/{filename}. For example, the CCAR-F track stores its diagnostic in ccar-f/diagnostic.json and mock exams in ccar-f/mock-01.json, with the track manifest located at certifications/claude/tracks/ccar-f.json.

Does the Python implementation validate assessment completeness?

The provided load_assessment() function loads JSON files but does not validate content completeness. It relies on the assessment_type() helper to categorize assessments based solely on the kind field, leaving comprehensive validation to the certification system's backend logic.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}