LOW-LAYER Automation

Claude Code automation configuration for development workflows: code standards, AI orchestration, custom commands, and security hooks.

This repository contains a production-ready Claude Code setup with workflows for release management, changelog generation, debugging, and more.

---

Features

• Security Hook: Pre-execution guard blocking dangerous operations (sudo, rm -rf, sensitive files)
• 11 Custom Commands: Specialized workflows for common development tasks
• Skills Auto-Detection: Context-aware skill loading via jeffallan/claude-skills
• Code & Testing Standards: Guidelines for code generation, documentation, and test pyramid
• AI Orchestration: Workflow patterns for effective Claude Code usage
• Task Management: File-based todo tracking and lessons learned system

---

Structure

low-layer-automation/
├── .claude/                         # Claude Code configuration
│   ├── settings.json                # Hooks and shared permissions
│   ├── settings.local.json          # Local permissions (not committed)
│   ├── hooks/
│   │   └── security-guard.sh        # Pre-execution security checks
│   ├── rules/                       # Workflow and task management
│   │   ├── WORKFLOW.md              # AI orchestration guidelines
│   │   ├── lessons.md               # Lessons learned and prevention rules
│   │   └── todo.md                  # Task templates and patterns
│   ├── tasks/                       # Active task plans (gitignored)
│   │   └── README.md                # Task directory documentation
│   ├── standards/                   # Externalized code standards
│   │   ├── code-standards.md        # File headers, comments
│   │   ├── readme-standards.md      # README template
│   │   ├── git-workflow.md          # Branching strategy
│   │   ├── changelog-standards.md   # Changelog format
│   │   └── testing-standards.md     # Test pyramid, per-stack conventions
│   ├── licenses/                    # License templates
│   │   ├── low-layer.license        # LOW-LAYER proprietary license
│   │   └── mit.license              # MIT license template
│   └── commands/                    # Custom slash commands
│       ├── changelog.md             # /changelog - Add changelog entries
│       ├── push.md                  # /push - Commit and push
│       ├── release.md               # /release - Full release workflow
│       ├── release-note.md          # /release-note - Generate release notes
│       ├── docker-release.md        # /docker-release - Build & push images
│       ├── debug.md                 # /debug - Systematic debugging
│       ├── epct.md                  # /epct - Explore-Plan-Code-Test
│       ├── arch-update.md           # /arch-update - Document architecture changes
│       ├── create-cmd.md            # /create-cmd - Create new commands
│       ├── rg.md                    # /rg - TDD bug fix (Red-Green)
│       └── web-design.md            # /web-design - UI/UX specialist
│
├── low-layer-architecture/           # Technical documentation (separate repo)
│   ├── platform.md                  # Platform architecture
│   ├── repositories.md              # Multi-repo structure
│   └── infrastructure.md            # Infrastructure architecture
│
├── CLAUDE.md                        # Project overview and standards index
└── .gitignore

---

Custom Commands

Command	Description
/changelog	Add entries to CHANGELOG.md based on current changes
/push	Commit and push with auto-generated message from changelog
/release	Complete release workflow (PR, merge, tag, GitHub release)
/release-note	Generate professional release notes from changelog
/docker-release	Build and push Docker image to ghcr.io
/debug	Systematic bug debugging with deep analysis
/epct	Explore-Plan-Code-Test workflow for features
/arch-update	Document architectural decisions and changes
/create-cmd	Create and optimize new slash commands
/rg	TDD bug fix: write a failing E2E test (Red), then fix the bug (Green)
/web-design	Web development and UI/UX specialist mode

---

Skills

Skills inject domain expertise (language patterns, framework conventions, testing strategies) into Claude Code sessions. Commands like /epct and /debug auto-detect relevant skills based on file extensions, project markers, and task keywords — no manual selection needed.

Skills are loaded from jeffallan/claude-skills (65 skills, 12 categories). Full documentation: jeffallan.github.io/claude-skills

Installation (Global)

This project uses global installation to avoid duplicating skills across repositories:

# Install the CLI
npx skills add https://github.com/Jeffallan/claude-skills --list   # Preview available skills

# Install skills globally (available across all projects)
npx skills add https://github.com/Jeffallan/claude-skills -g -s golang-pro
npx skills add https://github.com/Jeffallan/claude-skills -g -s angular-architect
npx skills add https://github.com/Jeffallan/claude-skills -g -s test-master
npx skills add https://github.com/Jeffallan/claude-skills -g -s kubernetes-specialist
npx skills add https://github.com/Jeffallan/claude-skills -g -s terraform-engineer
npx skills add https://github.com/Jeffallan/claude-skills -g -s typescript-pro

Global skills are stored in ~/.claude/skills/ and shared across all projects.

Recommended Profiles

Profile	Skills	Use case
Backend	golang-pro, postgres-pro, api-designer	API, CLI, Operator, Agent
Frontend	angular-architect, typescript-pro	Console (Angular 21)
Infra	kubernetes-specialist, terraform-engineer, devops-engineer	Operator, Helm, CAPI
Quality	test-master, code-reviewer, debugging-wizard	All projects

Management

npx skills ls -g                  # List globally installed skills
npx skills find <keyword>         # Search available skills
npx skills check                  # Check for updates
npx skills update                 # Update all skills
npx skills rm -g <skill-name>     # Remove a global skill

How Auto-Detection Works

Each skill has triggers in its frontmatter (e.g., .go, angular.json, Chart.yaml). When /epct or /debug runs:

1. Scans ~/.claude/skills/*/SKILL.md for installed skills
2. Matches triggers against the task description and involved files
3. Loads matched skills and injects their rules into all subagent prompts
4. Reports: "Skills loaded: golang-pro, test-master"

See WORKFLOW.md § Skill Auto-Detection for implementation details.

---

Security Hook

The security guard (.claude/hooks/security-guard.sh) blocks:

• Privileged commands: sudo, su, runas, chmod 777
• Destructive operations: rm -rf /, format, mkfs, dd
• System modifications: shutdown, reboot
• File access outside project: All Read/Write/Edit operations restricted
• Sensitive files: .env, credentials, API keys, SSH keys
• Dangerous git: git push --force, rm -rf .git
• Remote code execution: curl|bash, wget|sh, reverse shells

---

Quick Start

Prerequisites

• Claude Code: Latest version (claude.ai/code)
• Git: 2.30+
• GitHub CLI: 2.0+ (for /release and /docker-release commands)
• Bash: 4.0+ (Linux/macOS/WSL2)

1. Copy Configuration

# Copy the .claude directory and config files to your project
cp -r .claude /path/to/your/project/
cp CLAUDE.md /path/to/your/project/

2. Adapt Standards

Edit CLAUDE.md to match your project's:

• File header format
• Code commenting rules
• README structure
• Git workflow
• Changelog format

3. Customize Commands

Modify commands in .claude/commands/ to match your:

• Project paths
• Release workflow
• Docker registry
• Git branching strategy

---

Configuration Files

settings.json (Shared)

Hook configuration and shared permissions:

{
  "hooks": {
    "PreToolUse": [
      {
        "matcher": "Bash|Read|Write|Edit|Glob|Grep",
        "hooks": [
          {
            "type": "command",
            "command": "bash \"$CLAUDE_PROJECT_DIR/.claude/hooks/security-guard.sh\"",
            "timeout": 10,
            "statusMessage": "Security check..."
          }
        ]
      }
    ]
  },
  "permissions": {
    "allow": [
      "Bash(gh pr create:*)",
      "Bash(gh release create:*)"
    ]
  },
  "includeCoAuthoredBy": false
}

---

## Standards Overview

### CLAUDE.md

- **File Headers**: Mandatory copyright and purpose blocks
- **Section Headers**: Uppercase titles with descriptions
- **Comments**: Only for complex logic, not obvious code
- **README Format**: Comprehensive template with required sections
- **Git Workflow**: Version branches, feature branches, changelog-per-commit
- **Changelog Format**: Keep a Changelog + Semantic Versioning

### .claude/rules/WORKFLOW.md

- **Operating Principles**: Correctness, minimal changes, verification
- **Task Management**: File-based tracking in `.claude/tasks/`
- **Quality Assurance**: Test before commit, incremental delivery
- **Communication**: Concise, high-signal reporting

---

## Example Workflow

```bash
# 1. Start a new version
git checkout main && git pull
git checkout -b v1.2.0

# 2. Make changes...

# 3. Add changelog entry
/changelog low-layer-website

# 4. Commit and push
/push low-layer-website

# 5. More changes... repeat steps 2-4

# 6. Create release
/release low-layer-website

# 7. Build Docker image
/docker-release low-layer-website

---

Troubleshooting

Security hook blocks a legitimate operation

Cause: The pre-execution guard matches a pattern in your command.

Solution: Review .claude/hooks/security-guard.sh to understand the blocked pattern. If the operation is safe, temporarily disable hooks via /hooks menu or set "disableAllHooks": true in .claude/settings.json.

Skills not auto-detected

Cause: Skills are installed but not matched against the current task.

Solution: Verify skills are installed at ~/.claude/skills/<name>/SKILL.md. Check that the triggers field in the SKILL.md frontmatter matches your project markers (file extensions, technology names).

Command not found

Cause: Claude Code doesn't recognize the slash command.

Solution: Verify the command file exists at .claude/commands/<command>.md and follows the required format with argument-hint and allowed-tools fields.

---

License

This project is licensed under the MIT License — see the LICENSE file for details.

---

Author

LOW-LAYER Engineering - Toulouse, France

• GitHub: @Astocanthus
• Website: https://low-layer.com
• Contact: contact@low-layer.com

---

Acknowledgments

• Built for LOW-LAYER infrastructure automation
• Powered by Claude Code (Anthropic)
• Skills framework by jeffallan/claude-skills