Claude Code Development Platform

A battle-tested, distributable development environment for Claude Code + VS Code. Provides multi-agent safety, automated code review, deployment workflows, context conservation, and self-healing error handling — out of the box.

Born from production use across 3 repos and hundreds of development sessions. Packaged for any team to adopt and customize.

What You Get

Safety Rails (9 hooks)

• Branch protection — blocks commits to staging/main/master
• Worktree enforcement — agents work in isolated directories, can't overwrite each other
• Destructive operation blocking — no force push, hard reset, or checkout .
• PR merge prevention — agents create PRs, only humans merge
• Work registration — every agent registers their branch before committing
• Secrets detection — blocks commits containing .env, credentials, keys
• File redirect blocking — closes the echo > file worktree bypass
• Interactive mode blocking — prevents git rebase -i (no TTY in Claude Code)

Skills (15)

Skill	Purpose
/letsbuild	Create isolated worktree, assign window ID, register work
/shipit	Commit, push, create staging + production PRs
/code-review	Post-implementation code review with addon support
/security-review	OWASP-based security audit
/api-review	REST API design review
/pre-deploy	10-section deployment readiness check
/plan-eng-review	Architecture and engineering plan review
/plan-design-review	UX and design plan review
/plan-ceo-review	Founder-mode scope and strategy review
/retro	Weekly engineering retrospective from git history
/document-release	Post-ship documentation sync
/token-audit	Context efficiency analysis
/perf-test	Performance benchmarks and load testing
/feature	Feature request backlog management
/effort-optimizer	Task-based effort level and model recommendations

Agents (4)

Agent	Model	Purpose
code-reviewer	Opus	Deep code review with project-specific addons
deploy-verifier	Sonnet	Post-deployment health checks
perf-tester	Sonnet	Performance benchmarks and load testing
error-fixer	Sonnet	Self-healing config error diagnosis (one-try rule)

Smart Features

• Context conservation — tracks tool calls, warns at 50/80/120, suggests new windows
• Compaction protection — injects critical rules before context compression
• Self-healing errors — detects config errors, attempts one fix, escalates on failure
• Dynamic effort — recommends /effort and /model based on task complexity
• Permission presets — cautious, standard, power-user (reduces approval friction)
• Addon pattern — project-specific checks without modifying platform files

Quick Start

# 1. Clone or use as template
git clone https://github.com/TheLineExp/claude-code-platform.git
cd claude-code-platform

# 2. Run the setup wizard
bash setup.sh

# 3. Open Claude Code
# The wizard generates CLAUDE.md, settings.json, and platform.config.json

# 4. Start your first feature
/letsbuild

The wizard asks for your project name, branches, test command, deployment URLs, and feature tier — then generates everything.

Feature Tiers

Tier	Hooks	Skills	Agents	Best For
Minimal	All 9	/letsbuild only	None	Solo developers wanting safety rails
Standard	All 9	7 core skills	code-reviewer	Most teams (recommended)
Full	All 9	All 15 skills	All 4 agents	Teams wanting the complete platform

Permission Presets

Preset	Auto-Approves	Best For
Cautious	Reads only	New users, sensitive codebases
Standard	Reads, edits, common git/npm/docker	Most developers (recommended)
Power User	Almost everything (hooks are safety net)	Experienced users who want speed

Architecture

Your Repo/
├── platform.config.json          # Central config (branches, test cmd, URLs)
├── CLAUDE.md                     # Project rules (generated from template)
├── .claude/
│   ├── settings.json             # Permissions + hook bindings
│   ├── settings.local.json       # Personal overrides (gitignored)
│   ├── window-id                 # "setup" in main repo, "w1"/"w2" in worktrees
│   ├── active-work.md            # Agent work registry
│   ├── hooks/                    # 12 hook scripts
│   │   ├── _config.sh            # Shared config loader (reads platform.config.json)
│   │   ├── _parse-input.sh       # Shared JSON parser + fast-path optimization
│   │   ├── block-*.sh            # Safety hooks (5)
│   │   ├── check-*.sh            # Validation hooks (2)
│   │   ├── enforce-worktree.sh   # Worktree isolation
│   │   ├── session-tracker.sh    # Context conservation
│   │   ├── compact-notifier.sh   # Compaction protection
│   │   └── error-handler.sh      # Self-healing errors
│   ├── skills/                   # 15 skill definitions
│   └── agents/                   # 4 agent definitions
├── hooks/                        # Git hooks
│   ├── pre-commit                # Branch protection + secrets + tests
│   ├── pre-push                  # Production branch protection
│   ├── install.sh                # Hook installer
│   └── verify.sh                 # Drift detection
└── docs/                         # Documentation

How It Works

1. Hooks run on every tool call via settings.json bindings
2. Permissions auto-approve safe operations, deny dangerous ones
3. Skills provide structured workflows invoked with /skillname
4. Agents handle complex tasks (code review, deployment verification)
5. Config (platform.config.json) centralizes project settings — hooks read it at runtime
6. Addons (*-addons.md files) inject project-specific checks without modifying platform files

Customization

Project-Specific Checks (Addon Pattern)

Create addon files to extend platform skills:

<!-- .claude/code-review-addons.md -->
## Security (Critical)
- [ ] All PII encrypted with encryptionService
- [ ] Searchable fields have HMAC hashes

## Framework
- [ ] Business logic in services (not routes)
- [ ] Use asyncHandler wrapper on all Express routes

Skills automatically read their addon file if it exists.

Available Addon Files

File	Extends
.claude/code-review-addons.md	/code-review + code-reviewer agent
.claude/security-review-addons.md	/security-review
.claude/api-review-addons.md	/api-review
.claude/pre-deploy-addons.md	/pre-deploy
.claude/perf-test-addons.md	/perf-test
.claude/deploy-verifier-addons.md	deploy-verifier agent

See docs/CUSTOMIZATION.md for detailed customization guide.

Documentation

• SETUP.md — Detailed setup for each OS
• CUSTOMIZATION.md — How to add addons, skills, agents
• ARCHITECTURE.md — How the platform works internally
• SKILLS_REFERENCE.md — Every skill with usage examples
• TROUBLESHOOTING.md — Common issues and fixes

Examples

Pre-built configurations for common tech stacks:

• Node.js / Express
• Python / Django
• React Frontend
• FleetManager — The original production config

License

MIT