Claude Code

Why Claude Code?

Claude Code provides the most complete Fulcrum integration. While other vendors require flattening everything into single files or external references, Claude Code supports:

• Native slash commands — /start, /execute-work feel built-in
• @ mentions for agents — @tech-lead, @backend-engineer as first-class concepts
• Plugin-style skills — Invoke with /skill tool for structured workflows
• Symlinked structure — Clean separation of concerns, easy customization

Generated Structure

Running fulcrum sync --vendor claude generates a .claude/ directory with symlinked configuration:

.claude/
├── settings.json           # Claude Code configuration
├── settings.local.json     # Local overrides (gitignored)
├── commands/               # Slash commands (symlinks)
│   ├── start.md           → ../../fulcrum/commands/start.md
│   ├── execute-work.md    → ../../fulcrum/commands/execute-work.md
│   ├── status.md          → ../../fulcrum/commands/status.md
│   └── ...
├── agents/                 # Agent definitions (symlinks)
│   ├── tech-lead.md       → ../../fulcrum/agents/tech-lead.md
│   ├── backend-engineer.md → ../../fulcrum/agents/backend-engineer.md
│   └── ...
└── skills/                 # Skill plugins (symlinks)
    ├── brainstorming.md   → ../../fulcrum/skills/brainstorming.md
    ├── code-reviewer.md   → ../../fulcrum/skills/code-reviewer.md
    └── ...

Using Agents

Mention agents directly in your prompts using @ syntax:

# Get architectural guidance
@tech-lead Review this auth implementation for scalability concerns

# Implementation help
@backend-engineer Implement the caching layer we discussed

# Full council review
@tech-lead @backend-engineer @devops-engineer Review this PR together

Claude Code automatically loads the agent definition and adjusts its behavior accordingly. The agent's expertise, responsibilities, and decision-making framework become active context.

Available Agents

Using Slash Commands

Slash commands are native to Claude Code—type / and see available commands:

# Start a new work session
/start "Implement OAuth 2.0 with Google"

# Execute tracked work
/execute-work BEAD-001

# Check session status
/status

# Review your changes
/review

# Create a PR from your work
/pr

Core Commands

Using Skills

Skills are invoked via the /skill tool or inline mentions:

# Invoke skill directly
/skill brainstorming "Design a caching strategy"

# Skills auto-trigger from context
"Let's brainstorm the API design"  # Activates brainstorming skill

# Explicit skill with parameters
/skill code-reviewer --focus security --staged

Skill Categories

• Workflow Skills — brainstorming, writing-plans, executing-plans
• Quality Skills — code-reviewer, test-engineer, quality-gate-checker
• Review Skills — design-reviewer, council-review, pr-creator
• Specialized — documentation-writer, platform-architect

Symlink Fallback Behavior

When symlinks aren't available (Windows without developer mode, some network drives), Fulcrum falls back to file copying:

# Force copy mode
fulcrum sync --vendor claude --copy

# Auto-detect (uses symlinks if available)
fulcrum sync --vendor claude

Vendor Comparison

Configuration

Claude Code reads from .claude/settings.json and .claude/settings.local.json:

// .claude/settings.json (committed)
{
  "project": {
    "name": "my-project",
    "agents": "./agents",
    "commands": "./commands",
    "skills": "./skills"
  }
}

// .claude/settings.local.json (gitignored)
{
  "user": {
    "defaultAgent": "tech-lead",
    "autoSkills": true
  }
}