A professional-grade agentic system configuration for Claude Code, featuring RIPER workflows, specialized agents, automated hooks, and comprehensive quality gates.
- RIPER Workflow - Phase-separated development (Research, Innovate, Plan, Execute, Review)
- 6 Specialist Agents - Orchestration, code review, context management, dependency auditing
- 6 Automated Hooks - Skill suggestions, phase validation, quality gates, git safety
- Memory Bank - Persistent context across sessions
- Quality Gates - Automated code review, output validation, token budgeting
Option 1: Clone directly
git clone https://github.com/nategarelik/claude-agentic-config.git ~/.claudeOption 2: Run installer
# Unix/macOS
curl -fsSL https://raw.githubusercontent.com/nategarelik/claude-agentic-config/main/scripts/install.sh | bash
# Windows PowerShell
irm https://raw.githubusercontent.com/nategarelik/claude-agentic-config/main/scripts/install.ps1 | iexOption 3: Manual setup
- Clone or download this repository
- Copy contents to
~/.claude/ - Restart Claude Code
# Check hooks are configured
cat ~/.claude/settings.json | grep hooks
# Test a hook
echo '{"prompt": "fix bug"}' | python ~/.claude/hooks/skill-auto-activator.py| Agent | Purpose | Model |
|---|---|---|
workflow-orchestrator |
Routes tasks to appropriate skills/workflows | opus |
code-review-sentinel |
Automated code quality review | sonnet |
context-curator |
Manages and compresses session context | sonnet |
dependency-auditor |
Security and health analysis of dependencies | sonnet |
plugin-capability-scout |
Discovers plugins and MCP servers | sonnet |
agentic-system-architect |
Designs multi-agent systems | opus |
| Hook | Trigger | Purpose |
|---|---|---|
skill-auto-activator |
UserPromptSubmit | Suggests relevant superpowers skills |
riper-phase-validator |
PreToolUse | Enforces RIPER phase constraints |
session-archiver |
Stop | Archives session to memory bank |
token-budget-guardian |
PreToolUse | Monitors and limits token usage |
output-quality-gate |
SubagentStop | Validates output quality |
git-safety-net |
PreToolUse | Prevents dangerous git operations |
1. brainstorming -> Refine requirements
2. write-plan -> Create implementation plan
3. test-driven-dev -> Implement with TDD
4. code-review -> Automated review
5. verify-completion -> Final verification
1. systematic-debugging -> Investigate root cause
2. root-cause-tracing -> Trace to source
3. test-driven-dev -> Write regression test
4. verify-completion -> Confirm fix
| Phase | Mode | Allowed Actions |
|---|---|---|
| Research | Read-only | Grep, Glob, Read, WebSearch |
| Innovate | Read-only | + Brainstorming, design exploration |
| Plan | Read + Memory | + Write to memory-bank/ |
| Execute | Full access | All tools |
| Review | Read + Test | Run tests, code review |
Edit ~/.claude/settings.json to customize:
{
"hooks": {
"UserPromptSubmit": [...],
"PreToolUse": [...],
"Stop": [...]
}
}| Variable | Purpose | Default |
|---|---|---|
RIPER_PHASE |
Override current RIPER phase | execute |
TOKEN_BUDGET |
Session token limit | 200000 |
HOOK_LOG_LEVEL |
Hook logging verbosity | INFO |
~/.claude/
├── agents/ # Specialist agent definitions
│ ├── workflow-orchestrator.md
│ ├── code-review-sentinel.md
│ └── ...
├── hooks/ # Automated hook scripts
│ ├── skill-auto-activator.py
│ ├── riper-phase-validator.py
│ └── tests/ # Hook test suite
├── docs/ # Documentation
│ ├── guides/ # How-to guides
│ └── reference/ # API reference
├── memory-bank/ # Persistent context
│ └── main/
│ ├── plans/
│ ├── reviews/
│ ├── sessions/
│ └── decisions/
├── scripts/ # Installation scripts
├── CLAUDE.md # Token-optimized instructions
└── settings.json # Hook configuration
- Getting Started
- Architecture Overview
- Creating Agents
- Creating Hooks
- Workflow Guide
- Agent Reference
- Hook Reference
See CONTRIBUTING.md for guidelines.
MIT License - see LICENSE for details.
- Built with Claude Code
- Inspired by superpowers-marketplace
- Uses patterns from RIPER workflow and AB Method