---

name: thx-boris description: Production-tested patterns for maximizing Claude Code effectiveness. Covers CLAUDE.md setup, subagent design, hooks, permissions, parallel sessions, and MCP integration. Not for general coding or debugging. argument-hint: "[topic] e.g., 'hooks', 'CLAUDE.md', 'subagents', 'permissions', 'parallel sessions'" license: MIT compatibility: Claude Code CLI, Claude.ai, and web. Requires shell for scripts/. metadata: author: saad-ahmed version: 3.0.0 category: developer-tools tags: [claude-code, workflow, automation, devtools]

thx-boris: Claude Code Mastery

Production-tested patterns for maximizing Claude Code effectiveness, based on workflows from the Claude Code team.

Instructions

Step 1: Identify the User's Goal

Determine which workflow the user needs:

Goal	Section	Reference
Set up a new project	The Living CLAUDE.md Pattern	assets/claude-md-template.md
Create an agent	Subagent Design	references/subagent-templates.md
Configure hooks	Hook Automation	references/hooks-patterns.md
Optimize permissions/sessions/MCP	—	references/session-and-mcp.md
Run parallel sessions	Parallel Orchestration	—
Create your own skill	—	references/skill-authoring.md
Troubleshoot issues	—	references/troubleshooting.md
Avoid common mistakes	—	references/anti-patterns.md

Step 2: Apply the Relevant Pattern

Follow the pattern for the user's goal. Each pattern includes step-by-step instructions, success criteria, and common pitfalls.

Step 3: Validate the Setup

Run scripts/validate-skill.sh to verify configuration, or run scripts/setup-project.sh to bootstrap a new project.

---

The Living CLAUDE.md Pattern

Core principle: Every Claude mistake becomes a permanent lesson.

Structure

# Development Workflow

**Always use `[package-manager]`, not `[alternative]`.**

## Commands
[Ordered by frequency of use]

## Code Style
[Project-specific patterns Claude should follow]

## Anti-Patterns
[Things Claude got wrong - add new ones as discovered]

## Domain Knowledge
[Project-specific context Claude needs]

The Feedback Loop

Claude makes mistake → Human notices → Add to CLAUDE.md → Claude never repeats

Real-world example from Claude Code development:

## Anti-Patterns
- Don't use `enum` for union types — use `type Status = 'active' | 'inactive'`
  (Claude kept generating TypeScript enums until this was added)

Team CLAUDE.md Protocol

For team projects, CLAUDE.md is a shared artifact:

1. Check into git alongside code
2. All team members contribute
3. Review CLAUDE.md changes in PRs
4. Treat as living documentation

A ready-to-use template is available at assets/claude-md-template.md.

---

Subagent Design

Subagents are specialized Claude instances with focused responsibilities. Store in .claude/agents/.

When to Create Subagents

Task Type	Subagent?	Rationale
Repetitive validation	Yes	Consistent checks
Code review patterns	Yes	Domain expertise
Complex multi-step	Yes	Focused context
One-off tasks	No	Overhead not worth it

Running Skills as Subagents

Use context: fork in SKILL.md frontmatter to run a skill in an isolated subagent. The skill content becomes the subagent's task prompt (no conversation history).

---
name: deep-research
description: Research a topic thoroughly
context: fork
agent: Explore
---

Available agent types: Explore (fast codebase search), Plan (architecture), general-purpose (default).

Available Agent Templates

Ready-to-use agents in assets/agents/:

• build-validator - Verify builds pass before commit
• code-simplifier - Reduce complexity
• verify-app - End-to-end application testing
• dependency-updater - Safe dependency updates
• migration-runner - Database migration with rollback
• security-scanner - OWASP-based security checks

See references/subagent-templates.md for additional templates (code-architect, oncall-guide, pr-reviewer, test-writer).

Creating a Subagent

# Create agent file
touch .claude/agents/[agent-name].md

# Agent file structure:
# 1. Purpose (one line)
# 2. Trigger conditions
# 3. Step-by-step procedure
# 4. Success criteria
# 5. Handoff instructions

---

Hook Automation

Hooks run automatically before/after Claude actions. Configure in .claude/settings.json.

Common Patterns

Auto-format on write:

{
  "PostToolUse": [{
    "matcher": "Write|Edit",
    "hooks": [{
      "type": "command",
      "command": "bun run format || true"
    }]
  }]
}

Lint before commit:

{
  "PreToolUse": [{
    "matcher": "Bash(git commit*)",
    "hooks": [{
      "type": "command",
      "command": "bun run lint:claude && bun run test"
    }]
  }]
}

CRITICAL: PostToolUse hooks should always end with || true to avoid blocking Claude. PreToolUse hooks should NOT use || true so they block on failure.

See references/hooks-patterns.md for complete hook configurations including Notification, Stop, and chaining patterns.

---

Parallel Orchestration

Running multiple Claudes maximizes throughput for complex projects.

The 5-Tab Terminal Strategy (Boris's Pattern)

Tab 1: Claude — main coding session
Tab 2: Claude — research/exploration (isolated context)
Tab 3: git / manual commands
Tab 4: dev server / build watcher
Tab 5: test runner

This prevents context pollution between coding and research tasks.

Git Worktrees (Recommended)

# Create worktrees for parallel work
git worktree add ../myproject-feature-a feature-a
git worktree add ../myproject-feature-b feature-b

# Each worktree gets its own Claude instance
cd ../myproject-feature-a && claude
cd ../myproject-feature-b && claude

Task Decomposition

Bad: "Implement the entire authentication system"
Good: Split into parallel tracks:
  - Claude 1: Auth API endpoints
  - Claude 2: Auth UI components
  - Claude 3: Auth tests
  - Claude 4: Auth documentation

Custom Slash Commands & Skills

Create project-specific commands in .claude/commands/:

# .claude/commands/commit-push-pr.md
Commit all changes with a descriptive message, push to origin, and open a PR.

Design principles: Atomic, Idempotent, Verbose, Recoverable.

March 2026 skill features (for skill authors):

Feature	Description
context: fork	Run skill in isolated context
context: agent	Run as subagent, returns summary
$ARGUMENTS	User input after skill name
$CLAUDE_SKILL_DIR	Path to skill directory
!`command`	Dynamic context injection
allowed-tools	Restrict which tools skill can use
disable-model-invocation	Prevent auto-triggering
effort: low|medium|high	Hint at expected complexity

See references/skill-authoring.md for full documentation.

For permissions, sessions, and MCP configuration, see references/session-and-mcp.md.

---

Examples

Example 1: Setting up a new TypeScript project

User says: "Set up Claude Code for my Next.js project"

Actions:

1. Run scripts/setup-project.sh . pnpm
2. Copy assets/claude-md-template.md to CLAUDE.md
3. Customize CLAUDE.md with project-specific commands and style
4. Copy relevant agents from assets/agents/ to .claude/agents/

Result: Complete Claude Code configuration with hooks, permissions, and agents.

See templates/typescript-project/ for a working example.

Example 2: Creating a custom subagent

User says: "Create a subagent for code review"

Actions:

1. Consult references/subagent-templates.md for the pr-reviewer template
2. Copy to .claude/agents/pr-reviewer.md
3. Customize trigger conditions and review checklist

Result: Agent file ready to use in .claude/agents/.

Example 3: Configuring hooks for a Python project

User says: "Set up auto-formatting hooks for my Python project"

Actions:

1. Consult references/hooks-patterns.md for the Python project pattern
2. Create .claude/settings.json with Black + isort PostToolUse hooks
3. Add pytest + mypy PreToolUse hooks for pre-commit

Result: Auto-formatting on every file change, validation before every commit.

See templates/python-project/ for a working example.

---

Troubleshooting

Hooks not running

Cause: Invalid JSON in .claude/settings.json or wrong matcher pattern Solution: Validate JSON with cat .claude/settings.json | jq . and check matcher syntax in references/hooks-patterns.md

Claude ignores CLAUDE.md instructions

Cause: File not at project root, wrong filename, or instructions too verbose Solution: Verify file is exactly CLAUDE.md at root. Keep instructions concise with bullet points. Move details to references.

Skill doesn't trigger

Cause: Description too vague or missing trigger phrases Solution: Run scripts/validate-skill.sh to check. Add specific trigger phrases to the description field.

Over-triggering

Cause: Description too broad Solution: Add negative triggers ("Do NOT use for...") and be more specific about scope.

See references/troubleshooting.md for comprehensive troubleshooting.

---

Model Selection

Default: Opus 4.6 (the most capable Claude model).

Why Opus over Sonnet for complex work:

• Less steering required
• Better tool use
• Fewer mistakes = faster overall

When Sonnet 4.6 is acceptable:

• Simple, well-defined tasks
• High-volume, low-complexity work
• When latency matters more than quality

Override per-skill with model: in frontmatter.

---

Quick Reference

Task	Solution
Claude repeats mistake	Add to CLAUDE.md
Repetitive workflow	Create subagent in .claude/agents/
Auto-format code	PostToolUse hook
Reduce permission prompts	/permissions allow
Complex feature	Parallel Claudes with git worktrees
Common multi-step	Create a skill in .claude/skills/
Side-effect workflow	Skill with disable-model-invocation: true
Isolated research	Skill with context: fork + agent: Explore
New project setup	Run scripts/setup-project.sh

---

References

• references/subagent-templates.md - Complete subagent templates
• references/hooks-patterns.md - Hook configuration examples
• references/session-and-mcp.md - Sessions, permissions, MCP config
• references/skill-authoring.md - Creating your own skills
• references/troubleshooting.md - Common issues and fixes
• references/anti-patterns.md - What NOT to do
• assets/agents/ - Ready-to-use agent files
• assets/claude-md-template.md - Copy-paste starter for new projects
• templates/ - Complete TypeScript and Python project examples
• scripts/setup-project.sh - Bootstrap Claude Code for a new project
• scripts/validate-skill.sh - Validate skill folder structure
• scripts/validate-claude-md.sh - Validate CLAUDE.md structure