Exam Weight 20%
26 Articles · 6 Tasks
Claude Code Configuration
Task 3.1 Config Hierarchy (4)
K3.1.1
arrow_forward
Put Team Standards Where the Team Can Actually Get Them
Three-level config hierarchy
K3.1.2
arrow_forward
Mentioning a File Is Not the Same as Loading It
@import syntax
K3.1.3
arrow_forward
Twelve Unscoped Rules Files Waste 78% of Your Context Budget
.claude/rules/ modular rules
S3.1.1
arrow_forward
/memory Tells You What Is Actually Loaded — MEMORY.md Does Not
/memory command diagnosis
Task 3.2 Commands vs Skills (5)
K3.2.1
arrow_forward
Commands Are Prompts. Skills Are Configured Workflows.
Commands vs Skills
K3.2.2
arrow_forward
Three Fields That Turn a Prompt Into a Safe Workflow
SKILL.md frontmatter
K3.2.3
arrow_forward
Fork Isolates Memory, Not the Filesystem
context: fork isolation
K3.2.4
arrow_forward
Personal Skills Live in ~/.claude/skills/ — Name Them Differently
Personal skill customization
S3.2.1
arrow_forward
CLAUDE.md Is Always-On. Skills Are On-Demand. Do Not Mix Them Up.
Skills vs CLAUDE.md selection
Task 3.3 Frontmatter Paths & Globs (3)
K3.3.1
arrow_forward
86% of Your Rules Context Is Wasted Without Path Globs
YAML frontmatter paths glob
K3.3.2
arrow_forward
Section Headers Do Not Enforce Scope — Path Rules Do
Path rules reduce irrelevant context
K3.3.3
arrow_forward
File TYPE or File LOCATION — That Decides Which Mechanism to Use
Glob rules vs subdirectory CLAUDE.md
Task 3.4 Plan Mode & Exploration (4)
K3.4.1
arrow_forward
55% Rollback Rate When You Skip Planning on Complex Tasks
Plan mode for safe exploration
K3.4.2
arrow_forward
If You Already Know the Fix, Just Fix It
Direct execution for scoped changes
K3.4.3
arrow_forward
120 Files Explored, 12% Context Used
Explore subagent for context isolation
S3.4.1
arrow_forward
Plan → Explore → Direct: The Three-Phase Workflow for Complex Tasks
Plan + Explore + Direct combo
Task 3.5 Examples Over Descriptions (4)
K3.5.1
arrow_forward
Text Descriptions Hit an Ambiguity Ceiling at 63%. Examples Reach 91%.
Concrete examples > text descriptions
K3.5.2
arrow_forward
Tests First: 3 Iterations to 15/15. Implementation First: 7 Iterations to 12/15.
Test-driven iteration
K3.5.3
arrow_forward
Three Rounds of Rework Because Nobody Asked About Deployment Topology
Interview mode
K3.5.4
arrow_forward
Classify Before Submitting: Interacting Together, Independent Separately
Interacting vs independent issues
Task 3.6 CI & Non-interactive Mode (6)
K3.6.1
arrow_forward
Without -p, Your CI Job Hangs Until Timeout Kills It
-p flag for non-interactive CI
K3.6.2
arrow_forward
Prompt Says "Output JSON" → 73%. Schema Flag → 99%.
Structured CI output
K3.6.3
arrow_forward
CLAUDE.md Is Your CI Agent's Only Source of Project Standards
CLAUDE.md as CI project context
K3.6.4
arrow_forward
Claude Reviewing Its Own Code Is Barely a Review
Session isolation for unbiased review
S3.6.1
arrow_forward
62% of Review Comments Were Duplicates. Prior Context Dropped It to 4%.
Incremental review
S3.6.2
arrow_forward
60% Duplicate Test Suggestions Because Nobody Showed Claude the Existing Tests
Test generation dedup