// Track

Skills, Hooks & Slash Commands

Customizing the Claude Code Harness

The Claude Code harness is yours to shape — settings.json for mechanical enforcement, hooks for deterministic event reactions, skills for load-on-demand expert workflows, slash commands for versioned prompt shortcuts, and CLAUDE.md for semantic context. Master all five layers and the curator discipline that keeps them compounding.

Recommended: Complete Context Engineering & CLAUDE.md Architecture first

10 lessons~73 min total

Lessons are shown in recommended order. Complete them in sequence for the best experience — or jump to any lesson.

Lesson 383·7 min read

The Harness Is Yours to Shape

Claude Code's behavior is configurable at five layers — settings.json, hooks, skills, slash commands, and CLAUDE.md — and every layer you customize is a decision that fires automatically, every session, without touching your token budget.

Lesson 384·6 min read

Slash Commands — Reusable Prompts as Commands

A slash command is a prompt you've already debugged — saved as a markdown file, invoked with a word, and version-controlled with the project.

Lesson 385·8 min read

Skills — Packaging Expertise as .md Files

A skill is a multi-step expert workflow saved as a markdown file — the difference from a slash command is the verification gate: a skill knows when it's done.

Lesson 386·7 min read

Authoring a Skill the TDD Way

The /writing-skills approach inverts the usual authoring order: write the failure scenario first, then the minimal skill, then harden it with rationalization counters — because skills built from failure handle failure.

Lesson 387·8 min read

Hooks — Deterministic Enforcement

Hooks intercept the agent loop at three injection points — PreToolUse, PostToolUse, and Stop — using exit codes to block, warn, or allow. Unlike CLAUDE.md rules, hooks fire deterministically regardless of context state.

Lesson 388·7 min read

Hook Architecture & Security

Hooks run in the critical path of every tool call — a bug in a hook silently breaks the agent loop. Understand what hooks can do, what they can't, and how to guard the blast radius of each one you deploy.

Lesson 389·6 min read

Permission Modes & the Allowlist

Four permission modes control how aggressively Claude Code asks for confirmation — and an allowlist designed correctly eliminates 90% of prompts without sacrificing any safety.

Lesson 390·7 min read

Model Routing & Subagent Config

Model routing is the cost dial — Sonnet for 90% of tasks, Opus where reasoning depth changes the outcome. Subagent config adds tool restrictions that limit blast radius per agent role.

Lesson 391·8 min read

Building a Skill Library

Sixty-plus skills organized by domain, curated monthly, with triggers that fire reliably — the difference between a skill library that compounds and one that accumulates dead weight.

Lesson 392·9 min read

The Customization Playbook

Assembling settings, hooks, skills, and commands into a coherent personal harness — and the monthly audit lab that keeps every layer healthy, calibrated, and earning its place.