\ | /
                                                       --\*/--
                          ___                            / | \
                         /   \
                        | ^ ^ |
                         \_u_/
                      _____|_____
          .---.     /     |     \     .---.
         | O  |---/   /---|---\  \---| O  |
         | /\ |  |   /    |    \  |  | /\ |
         |/  \|  |  |     |     | |  |/  \|
         '----'  |  |     |     | |  '----'
                  \  \    |    /  /
                   \  ----|---  /
                    |    / \   |    how Claude perceives me
                    |   /   \  |    
                    |  /     \ |
                    | /       \|
                    |/         '
                    '

Claude Code Day 2 Workshop (2026 Edition)

A hands-on workshop for developers learning Claude Code's most powerful features. 11 katas total, ~3 hours if you pick 4–5 of them. Each kata has a quick Fundamentals section for newcomers and a meatier Novel Layer (2026) section that surprises even daily users.

Who Is This For?

Two audiences side by side:

• Newcomers — have tried Claude Code once or twice, want the foundation
• Daily users — want to see what's changed in the 2026 ecosystem (plugins, 27-event hooks, per-agent model pinning, elicitation, statusline, headless workflows)

Both should leave with new tricks.

Workshop Schedule (~3 hr)

Facilitators pick 3–4 katas based on audience interest. Each takes 20–35 min including a brief demo.

#	Kata	Time	Fundamentals	Novel layer (2026)
0	Prompting Patterns	Reference	Universal patterns across Claude Code, Claude.ai, Cursor, Copilot	Spec-driven dev, effort levels, prompt-as-judge, backpressure
0	Directory & Config Files	Reference	CLAUDE.md, .claude/, settings layers	Statusline, agents/, plugins/, env-based model routing
00	CLAUDE.md & Sub-Agents Example	Reference	Layered CLAUDE.md, sub-agent dispatch patterns	YAML-frontmattered agents, per-agent model pinning
01	Agent Modes	~20 min	Permission modes, Shift+Tab, Plan mode	/effort, background mode (&), Agent View, claude -p
02	Security & Containment	~25 min	Permission rules, /sandbox, file boundaries	.claudeignore, prompt-type hook as AI judge, HTTP hooks
03	Variables & Statusline	~25 min	Settings layers, CLAUDE.md, env vars	Custom statusline (context %, cost, model), CLAUDE_CODE_SUBAGENT_MODEL routing
04	Hotkeys & Slash Commands	~20 min	Default shortcuts, Vim mode, multiline	/effort, /agents, /plugin, /handoff, custom keybindings
05	Skills & Plugin Marketplaces	~30 min	SKILL.md, /command, dynamic ! injection	Progressive disclosure (refs/scripts/examples), /plugin install, building a private marketplace
06	Hooks Mastery	~30 min	PreToolUse/PostToolUse, exit codes, matchers	27+ events (UserPromptSubmit, SessionStart, PreCompact), 4 handler types (command/http/prompt/agent), async hooks
07	Custom MCP Server with Elicitation	~35 min	stdio server, tools/resources, registration	Elicitation (server asks user mid-task via form), packaging MCP as a plugin
08	Expert Topics	Reference	(none — pure reference)	Agent Teams, Agent View, Remote Control, headless CI patterns, token economics

Suggested 4-Kata Tracks

"Showing off what 2026 brought" — for daily users who want the wow:

03 (Statusline) → 05 (Skills + Plugins) → 06 (Hooks) → 07 (MCP + Elicitation)

"Practical foundations" — for mixed audience where some are new:

01 (Agent Modes) → 03 (Variables + Statusline) → 05 (Skills + Plugins) → 06 (Hooks)

"Enterprise/security focus" — for regulated environments:

02 (Security) → 03 (Variables) → 06 (Hooks) → 08 (Expert: governance, headless)

Prerequisites

• Claude Code installed and authenticated (claude command works). Latest version recommended (npm install -g @anthropic-ai/claude-code).
• Node.js 18+ (for kata 07)
• Git
• jq available on the PATH (used by hook scripts in 02, 03, 06)
• A terminal (macOS, Linux, or Windows via PowerShell / Git Bash)

Cross-Platform Support

All katas include setup instructions for macOS/Linux (bash) and Windows (PowerShell). On Windows, statusline and hook scripts run through Git Bash if installed (use forward slashes in paths) or PowerShell directly.

How to Use

1. Clone this repo
2. Facilitator picks 4–5 katas to cover live
3. Self-paced learners work through the rest after the workshop

Each kata has:

• Theory — Fundamentals (2–3 min read, skip if you already know it)
• Theory — Novel Layer (2026) (3–5 min read, this is where the value is)
• Exercise — Fundamentals (5–10 min hands-on)
• Exercise — Novel Layer (10–20 min hands-on)
• Discussion Points — team conversation starters

Running the Workshop

Self-paced:

• Each kata is self-contained with its own setup/teardown
• Recommended order: 01 → 03 → 05 → 06 → 07 → 08 if working through everything