Universal Claude Code Workspace v2

Orchestration framework for Claude Code with 15 skills, 10 agents, 9 commands, and production hooks.

Quick Start

# Setup
& "C:\Program Files\Git\bin\bash.exe" setup.sh

# Start
claude

Skills (15)

Category	Skill	Triggers
Backend	fastapi-crud	endpoint, API, CRUD, model
	database	migration, PostgreSQL, pgvector, Alembic
	auth	JWT, OAuth, login, RBAC, password
	caching	Redis, cache, rate limiting, pub/sub
	websocket	WebSocket, real-time, chat, broadcast
	api-design	REST, versioning, pagination, errors
	monitoring	logging, metrics, tracing, health check
Frontend	frontend-design	React, Vue, Tailwind, component
DevOps	docker	Dockerfile, container, compose
	ci-cd	GitHub Actions, pipeline, deploy
Security	vuln-analysis	vulnerability, CVE, CWE, injection
	exploit-dev	PoC, exploit, payload
	patch-gen	patch, fix, remediation
	gui-automation	desktop automation, Agent S
Shared	ethics + vuln-taxonomy	referenced by security skills

Agents (10)

User request → @pm-spec → @architect → @coder → @reviewer → @tester

Agent	Model	Role
pm-spec	opus	Requirements → specs
architect	opus	System design → ADRs
coder	sonnet	Implementation
reviewer	opus	Code review (read-only)
tester	sonnet	QA + test automation
designer	sonnet	UI/UX design
devops	sonnet	Infrastructure + deploy
researcher	opus	Technical research + web search
offsec-specialist	opus	Security testing (RAPTOR)
gui-automation-specialist	opus	Desktop automation

Commands (9)

Command	Description
/plan	Plan feature with architect
/review	Multi-agent code review
/scan	Security scan (RAPTOR)
/exploit	Generate PoC
/patch	Generate security fix
/fuzz	Binary fuzzing (AFL++)
/forensics	GitHub forensic analysis
/deploy	Guided deployment
/handoff	Save session context

Hooks (7 lifecycle events)

Event	Hook	Purpose
PreToolUse:Bash	pre_bash_firewall.sh	Block destructive + security tools
PreToolUse:Edit	pre_edit_protect.sh	Protect sensitive files
PreToolUse:PR	pre_pr_check.sh	Tests + lint before PR
PostToolUse:Edit	post_edit_format.sh	Auto-format (black, prettier)
PostToolUse:Bash	post_bash_log.sh	Command logging
PostToolUse:Bash	post_scan_raptor.sh	RAPTOR tier loading
Stop	stop_validate.sh	Pre-completion validation
SubagentStop	subagent_checkpoint.sh	Auto-checkpoint (opt-in)
SessionStart	session_start_context.sh	Load session context
UserPromptSubmit	codebase_map.sh	Inject codebase structure
ContextUpdate	context_warning.sh	Context usage warnings

All hooks share lib/common.sh library for deduplication.

Security Mode (RAPTOR)

export SECURITY_MODE=1

Commands: /scan → /exploit → /patch Personas: "Use Mark Dowd persona", "Use Pentester persona"

Structure

.claude/
├── agents/          # 10 agent definitions + README
├── commands/        # 9 slash commands
├── skills/          # 15 domain skills
│   ├── shared/      # ethics.md, vuln-taxonomy.md
│   └── <name>/      # SKILL.md + references/ + scripts/
├── hooks/
│   ├── hooks.json   # 7 lifecycle events
│   └── scripts/     # 11 hook scripts + lib/common.sh
└── settings.json    # permissions, env, hooks config
docs/
├── GOTCHAS.md       # Self-improving failure patterns
tiers/
├── analysis-guidance.md
├── recovery.md
└── personas/        # 8 security expert personas

Contributing

See CONTRIBUTING.md.