๐Ÿค– CCDV-F Study PDF Resources

Official Udemy Course Companion Materials โ€” Open locally from your Job Tracker repo

53
Questions
120
Minutes
72%
Pass Score
Dec 6
Exam Date
11
Weekends Left

๐Ÿค– CCDV-F Free Study & Local Hands-On Blueprint

Comprehensive 1,049-Line Study Manual covering the 8-Hour freeCodeCamp Syllabus by Andrew Brown, Zero-Cost Local CLI Architecture, and $5 Console Credit Budgeting.
โญ Live Blueprint in Git ๐Ÿ’ฐ Zero-Cost Local Framework ๐ŸŽฅ 8-Hour freeCodeCamp (8 Domains) ๐Ÿ’ต $5 Credit Budget Lab
๐Ÿ“– View Full Manual on GitHub ๐Ÿ“„ Open Raw Markdown
D1: Agents & Loops 15%
Orchestrator-Worker patterns, ReAct loops, Claude Agent SDK query() hooks, PreToolUse decision gate (allow/deny/ask).
D2: Messages & SSE 33%
Messages API payload spec, SSE event lifecycle (message_start โ†’ delta โ†’ message_stop), ephemeral prompt caching (0.1x read).
D3: Claude Code CLI Local
CLI architecture, workspace discovery, git dirty-state ingestion, .claude/config.json, compaction sliding window.
D4: Error Taxonomy Eval
Transient HTTP 429/529 (jitter backoff) vs Reasoning/Context token overflows vs Tool execution failures (is_error: true).
D5: Sampling Bounds Controls
Temperature (0.0โ€“1.0), Nucleus Sampling (top_p), Adaptive Extended Thinking (effort: medium/high) vs Legacy (budget < max).
D6: XML Tag Engine Content
Semantic isolation with <context>, <rules>, <scratchpad>, system message separation against prompt bleeding.
D7: Security Guardrails Safety
Direct/Indirect prompt injection defenses, client-side PII regex masking vault, blocking destructive commands (rm -rf /).
D8: Tools & MCP 30%
JSON Schema 2020-12, stop_reason: tool_use loops, MCP Primitives (Tools, Resources, Prompts), stdio vs HTTP transport.
๐Ÿ› ๏ธ Zero-API-Cost Local Workstation Commands:
# 1. Install official Claude Code CLI npm install -g @anthropic-ai/claude-code # 2. Local diagnostic check (Zero API tokens consumed) claude doctor # 3. Production CLAUDE.md Hierarchy Precedence: ~/.claude/CLAUDE.md (User global) โ””โ”€โ”€ ./CLAUDE.md (Repository root rules) โ””โ”€โ”€ ./services/api/CLAUDE.md (Subtree scoped) # 4. Local Stdio MCP Server Execution: python mcp_local_server.py # Communicates via stdin/stdout JSON-RPC 2.0
๐Ÿ’ต Anthropic Console $5 Credit Optimization Parameters:
โ€ข Model: claude-3-5-haiku-20241022 ($0.80/MTok Base In | $4.00/MTok Out) โ€ข Max Yield: Over 6.25M base input tokens or 62.5M cached read tokens! โ€ข Prompt Caching: 90% discount on cache hits (0.1x multiplier = $0.08 / MTok) โ€ข Cache Writes: 5-Min TTL (1.25x) vs 1-Hour TTL (2.0x) | Exactly 4 breakpoints max โ€ข Batch API: 50% discount on non-urgent requests (/v1/messages/batches) โ€ข Validation Regex: ^[a-zA-Z0-9_-]{1,64}$ for custom_id batch constraints
๐Ÿ“‹ Quick References
๐Ÿค–

CCDV-F Foundations Blueprint 1,049 lines

Official freeCodeCamp 8-Hour Syllabus tracker, Zero-Cost Local Engineering Framework (Claude CLI, CLAUDE.md hierarchy, stdio MCP Server), and $5 Free Credit Budget Optimization Laboratory.

โญ Live Blueprint 100% Free Tier
โšก

Cram Sheet 6 pages

All key terms, definitions, API fields, cost multipliers, and SSE sequence. Review this every weekend.

โœ… Reviewed
๐Ÿ“–

Study Companion 37 pages

Full textbook reference. Covers all 4 macro-domains, exam tips, and scenario walkthroughs by Jacob Bushong.

โœ… Reviewed
๐Ÿ“…

30-Day Study Plan 2 pages

6 practice exams phased across 30 days. Maps to our 11-weekend schedule. Print and post on wall.

โœ… Reviewed
๐Ÿ—‚๏ธ Domain Slide Decks
๐Ÿค–

Domain 1 โ€” Agents & Workflows 15%

182 slides. Agent loop, orchestrator-worker, context isolation, hooks, LangGraph, Strands, PydanticAI, Claude Agent SDK query().

โœ… W1โ€“W2 (Sep 20โ€“28)
๐Ÿ”Œ

Domain 2 โ€” Applications & Integration 33%

370 slides. Messages API, SSE streaming, prompt caching, message batches, Bedrock Mantle, Vertex AI. Heaviest exam domain.

โฌœ W3โ€“W6 (Oct 3โ€“25)
๐Ÿ’ป

Domain 3 โ€” Claude Code

CLAUDE.md hierarchy, Rules, Skills, Commands, Agents. Slash commands, session management, compaction hooks.

โฌœ W6 (Oct 24โ€“25)
๐Ÿงช

Domain 4 โ€” Eval, Testing & Debugging

Error classification (transient vs reasoning vs tool), re-prompt strategies, eval harnesses, debug tooling.

โฌœ W7 (Oct 31โ€“Nov 1)
โš™๏ธ

Domain 5 โ€” Model Selection & Optimization

Haiku/Sonnet/Opus trade-offs, sampling (temperature, top_p, top_k), effort levels, cost calculation.

โฌœ W7โ€“W8 (Oct 31โ€“Nov 8)
โœ๏ธ

Domain 6 โ€” Prompt & Context Engineering

Structured prompts, XML delimiting, system vs user placement, validate-and-reprompt, context window management.

โฌœ W8โ€“W9 (Nov 7โ€“15)
๐Ÿ›ก๏ธ

Domain 7 โ€” Security & Safety

Prompt injection defense, PII prevention, identity validation, destructive action blocking with hooks, monitoring.

โฌœ W9โ€“W10 (Nov 14โ€“22)
๐Ÿ”ง

Domain 8 โ€” Tools & MCPs 30%

Second-heaviest domain. Tool definitions, JSON Schema, client-side vs server-side tools. MCP primitives (Tools/Resources/Prompts), stdio vs HTTP transports, auth, roots. Authoring & registering MCP servers.

โฌœ W10โ€“W11 (Nov 21โ€“29)

โšก Quick Reference โ€” Cram Sheet Key Values

Concept Value / Pattern
Pass score 720/1000 (~72%) โ€” Pearson VUE proctored
Cache read cost 0.1x โ€” 90% discount
Cache write (5 min TTL) 1.25x
Cache write (1 hr TTL) 2.0x
Max cache breakpoints 4 per request
Message Batch discount 50% off
Batch custom_id pattern ^[a-zA-Z0-9_-]{1,64}$
Vertex AI required body field anthropic_version: "vertex-2023-10-16"
Adaptive thinking effort field output_config.effort โ†’ low/medium/high/max
SSE sequence message_start โ†’ content_block_start โ†’ input_json_delta โ†’ content_block_stop โ†’ message_stop
PreToolUse output field permissionDecision: allow | deny | ask
tool_choice values auto | any | tool | none