ring/docs/platforms/claude-code.md

Ring for Claude Code

Claude Code is the native platform for Ring. All skills, agents, commands, and hooks are designed primarily for Claude Code and require no transformation.

Installation

# Interactive installer
./installer/install-ring.sh

# Direct install
./installer/install-ring.sh install --platforms claude

# Via curl (remote)
curl -fsSL https://raw.githubusercontent.com/lerianstudio/ring/main/install-ring.sh | bash

Installation Path

Ring installs to: ~/.claude/

~/.claude/
├── plugins/
│   └── ring/                    # Ring marketplace
│       ├── default/             # Core plugin
│       │   ├── skills/          # 20 core skills
│       │   ├── agents/          # 5 specialized agents
│       │   ├── commands/        # 6 slash commands
│       │   └── hooks/           # Session lifecycle hooks
│       ├── dev-team/            # Developer agents plugin
│       ├── pm-team/             # Product planning plugin
│       ├── finops-team/         # FinOps plugin
│       └── tw-team/             # Technical writing plugin
└── .ring-manifest.json          # Installation tracking

Components

Skills (46 total)

Skills are markdown files with YAML frontmatter that define workflows, processes, and best practices. They auto-load at session start.

Invocation:

Skill tool: "test-driven-development"
Skill tool: "systematic-debugging"

Structure:

---
name: skill-name
description: |
  What this skill does  
trigger: |
  - When to use this skill  
skip_when: |
  - When NOT to use  
---

# Skill Content
...

Agents (20 total)

Agents are specialized AI personas with defined expertise and output schemas.

Invocation:

Task tool with subagent_type="code-reviewer"
Task tool with subagent_type="backend-engineer-golang"

Categories (invoke as ring-{plugin}:{agent-name}):

• Review agents: code-reviewer, business-logic-reviewer, security-reviewer
• Planning agents: write-plan, codebase-explorer
• Developer agents: 9 language/role specialists (*)
• FinOps agents: finops-analyzer, finops-automation
• TW agents: functional-writer, api-writer, docs-reviewer

Commands (14 total)

Slash commands provide quick access to common workflows.

Examples:

/codereview          # Parallel 3-reviewer dispatch
/brainstorm          # Socratic design refinement
/pre-dev-full        # 8-gate planning workflow

Hooks

Hooks run automatically at session events:

• SessionStart: Load skills reference, inject context
• UserPromptSubmit: Remind about CLAUDE.md
• Stop: Ralph Wiggum loop continuation

Session Lifecycle

1. Session Start: Ring hooks load, skill reference generated
2. Context Injection: Plugin introductions (using-ring, using-dev-team, etc.)
3. Skill Discovery: Check applicable skills before any task
4. Workflow Execution: Use skills, agents, commands as needed
5. Verification: Evidence-based completion

Configuration

No additional configuration required. Ring auto-discovers plugins from ~/.claude/plugins/ring/.

Optional: Plugin Selection

To install specific plugins only:

./installer/install-ring.sh install --platforms claude --plugins default,dev-team

Exclude Plugins

./installer/install-ring.sh install --platforms claude --exclude finops-team,pm-team

Updating

# Check for updates
./installer/install-ring.sh check

# Update all plugins
./installer/install-ring.sh update

# Sync only changed files
./installer/install-ring.sh sync

Uninstalling

./installer/install-ring.sh uninstall --platforms claude

Troubleshooting

Skills not loading

1. Check hooks are enabled: Settings → Hooks
2. Verify installation: ./installer/install-ring.sh list
3. Check hook output: Look for "Ring skills loaded" at session start

Agent not found

1. Ensure plugin is installed that contains the agent
2. Use fully qualified name: backend-engineer-golang

Commands not available

1. Commands require their parent plugin
2. Check /help for available commands

Best Practices

1. Always check skills first: Before any task, verify if a skill applies
2. Use parallel reviews: Dispatch all 3 reviewers simultaneously
3. Follow TDD: Test → Fail → Implement → Pass → Refactor
4. Evidence before claims: Run verification, show output