@adel13/claude-harness

claude-harness

Production-grade Claude Code configuration — hooks, rules, agents, and MCP templates.

Quick Start

npx claude-harness
# Restart Claude Code
claude-harness doctor

What's Included

Layer	Contents	Count
hooks	Pre-commit gates, security monitoring, session continuity, cost tracking	34 scripts
rules	Coding standards (TypeScript, Python, Go, Rust, Java, Kotlin, Swift, PHP, Perl, C++, C#)	75 files
agents	Specialized agents (planning, review, TDD, security, architecture, build fixing)	36 agents
mcp	MCP server templates (GitHub, Exa, Playwright, Firecrawl, Ollama, and more)	27 servers

Installation

npx (recommended)

npx claude-harness

Selective layers

npx claude-harness --layers hooks,rules

Full autonomous mode

npx claude-harness --profile jarvis

Standalone shell script

curl -fsSL https://raw.githubusercontent.com/AdelElo13/claude-harness/main/setup.sh | bash

CLI Commands

Command	Description
claude-harness	Interactive install (default)
claude-harness doctor	Check system health and installed layers
claude-harness update	Update installed layers to latest
claude-harness uninstall	Remove installed layers
claude-harness list	List installed layers and files
claude-harness backup	Backup current ~/.claude configuration
claude-harness add	Add a layer to an existing install

Options

Flag	Description
--layers=<list>	Comma-separated: core,hooks,rules,agents,mcp
--profile=<name>	CLAUDE.md profile: minimal, standard, jarvis
--claude-dir=<path>	Target directory (default: ~/.claude)
--yes	Skip confirmation prompts

CLAUDE.md Profiles

Profile	Description
minimal	Lean setup — core settings and basic instructions only. Good for exploratory use.
standard	Balanced configuration with rules, agent routing, and session continuity. Recommended for most developers.
jarvis	Full autonomous mode — all hooks active, broad permissions, multi-agent orchestration, Telegram remote control.

Plugin Tiers

Essential (auto-installed)

Installed automatically with the core layer. Required for hooks, agents, and rules to work correctly.

• superpowers — plan/execute workflow
• context7 — library documentation lookup
• commit-commands — conventional commits
• code-review — inline code review
• code-simplifier — refactor and simplify
• feature-dev — TDD feature workflow
• pr-review-toolkit — pull request review
• claude-md-management — CLAUDE.md editing
• hookify — hook configuration
• plugin-dev — plugin scaffolding

Recommended (manual)

Install individually with claude plugin install <name>:

Category	Plugins
Frontend	frontend-design, playground
Security	security-guidance
Language servers	typescript-lsp, pyright-lsp, swift-lsp
Communication	telegram
Development	agent-sdk-dev, skill-creator, ralph-loop
Output styles	explanatory-output-style, learning-output-style

Requirements

• Node.js >= 18
• Claude Code (claude.ai/code)
• Docker — optional, used by some hooks
• Python 3 — optional, used by security monitor hook

Platform Support

Platform	Status
macOS	Supported
Linux	Supported
WSL2	Supported
Windows (native)	Not supported

FAQ

Will it overwrite my existing config? No. The installer backs up your existing settings.json before making any changes and uses a merge strategy that preserves your customizations.

Can I uninstall? Yes. Run claude-harness uninstall to remove installed layers. Your original settings backup is preserved.

Where are credentials stored? Credentials are never written to config files. MCP servers are configured to read from environment variables. See the MCP layer templates for the expected variable names.

Can I install only specific layers? Yes. Use --layers=hooks,rules to install a subset. The core layer is always recommended as a base.

How do I add a language ruleset after install? Re-run npx claude-harness install --layers=rules and select the additional languages, or use claude-harness add.

License

MIT — see LICENSE.