Claude Code Template

An opinionated, comprehensive Claude Code starter template — installable as a plugin, forkable as a base repo, runnable in a devcontainer.

Covers: slash commands, subagents, auto-triggered skills, output styles, status line, hooks, MCP servers, CI/CD (GitHub Actions + GitLab), Agent SDK examples, devcontainer with egress firewall, and more.

Install as a plugin

If you already use Claude Code:

/plugin install scotthavird/claude-code-template

This gives you all commands, agents, skills, output styles, and hooks from this template, without forking the repo.

Fork as a base

1. Use this template to create a new repo.
2. Open in a container — VS Code / Cursor will prompt to "Reopen in Container", or use Cmd+Shift+P → Dev Containers: Reopen in Container. Claude Code, Prettier, ESLint, Ruff, and GH extensions are pre-installed.
3. Authenticate — copy .claude/settings.local.json.example → .claude/settings.local.json and add your ANTHROPIC_API_KEY, or run claude /login.
4. Start a session — claude in the repo root.
5. Try a command — /analyze-project, /review, or /debug.

What's inside

Slash commands (12)

/analyze-project /commit /pr /test /lint-fix /review /security-review /debug /refactor /explain /doc /implement-issue

Subagents (7)

security-auditor doc-generator test-runner pr-reviewer refactor-planner debugger dependency-auditor

Skills (7, auto-triggered)

code-review db-migration test-writing api-design performance-audit accessibility effort-aware

Output styles (3)

concise educational review

Hooks (real, not just logging)

• Format on save — Prettier / Ruff / gofmt / rustfmt
• Block dangerous bash — destructive patterns, force-push to main, pipe-to-shell (with quote/comment-aware matching)
• Inject context on session start — branch, commits, open PRs
• Redact secrets from tool output — uses v2.1.122 updatedToolOutput to scrub keys/JWTs/PATs before the model sees them
• PreCompact checkpoint — saves session state before compaction drops context (v2.1.105+)
• Session cost on stop — total cost plus per-tool duration_ms breakdown (v2.1.121+)

CI/CD

• .github/workflows/claude.yml — @claude mentions in issues/PRs
• .github/workflows/claude-review.yml — automatic review on new PRs
• .gitlab-ci.yml — GitLab equivalent
• scripts/ci-review.sh — headless-mode example

Agent SDK

sdk/ has TypeScript and Python starters including a custom-tool example.

DevContainer

• Node 20 + Python 3.12 + Docker-in-Docker
• Egress firewall (init-firewall.sh) restricting outbound network to an allowlist (Anthropic API, GitHub, npm, PyPI)
• ~/.claude mounted so global settings and auto-memory persist

Directory layout

.claude-plugin/      plugin.json + marketplace.json
.claude/             commands, agents, skills, output-styles, statusline, settings
.devcontainer/       container + firewall
.github/workflows/   Claude Code CI
.mcp.json            filesystem, memory, git, fetch
docs/                best-practices, permission-modes, hooks-cookbook, plugins, agent-sdk, integrations
scripts/             hook scripts + log analyzer + CI review
sdk/                 TypeScript + Python SDK starters
CLAUDE.md            persistent system prompt

Documentation

• Best practices
• Permission modes
• Hooks cookbook
• Plugins
• Agent SDK
• Integrations
• The .claude/ directory

Customization

Add a slash command

Create .claude/commands/my-command.md:

---
description: What this command does
argument-hint: [arg1]
allowed-tools: Bash(git:*), Read
---

Your prompt here. $ARGUMENTS becomes the user's input.

Add a subagent

Create .claude/agents/my-agent.md:

---
name: my-agent
description: When this agent should be used.
tools: Read, Grep, Glob, Bash
model: sonnet
---

System prompt for the agent...

Add a skill

Create .claude/skills/my-skill/SKILL.md:

---
name: my-skill
description: Triggers on X, Y, Z contexts.
allowed-tools: Read, Grep
---

Instructions for the skill...

Security notes

• logs/ is gitignored — hook logs may contain sensitive prompts/outputs.
• .claude/settings.local.json is gitignored — put personal API keys there.
• Permission rules deny Read(.env), Read(./secrets/**), Bash(rm -rf:*), Bash(sudo:*), and pipe-to-shell patterns by default.
• The PreToolUse hook block-dangerous-bash.sh provides a second layer.
• The devcontainer firewall restricts egress to a fixed allowlist.

Learn more

• Claude Code overview
• Commands
• Hooks
• Skills
• Subagents
• Plugins
• Settings
• MCP
• Agent SDK
• GitHub Actions
• DevContainers

Contributing

Fork, customize, share. If you add a broadly useful command / skill / agent, PRs are welcome.

License

MIT — see LICENSE.