start

Session Startup Health Check

Run a comprehensive environment and configuration health check, then provide a session startup checklist.

Process

Step 1: Core Environment

Run these bash commands to gather environment info:

# Claude Code version
claude --version 2>/dev/null || echo "claude CLI: not found"

# Runtime versions
node --version 2>/dev/null || echo "node: not found"
python3 --version 2>/dev/null || echo "python3: not found"
git --version 2>/dev/null || echo "git: not found"
gh --version 2>/dev/null | head -1 || echo "gh: not found"

Step 2: GitHub Authentication

gh auth status 2>&1 | head -5

Step 3: MCP Servers

Check which MCP servers are configured. The primary config is ~/.claude/mcp.json:

cat ~/.claude/mcp.json 2>/dev/null | python3 -c "
import sys, json
d = json.load(sys.stdin)
servers = d.get('mcpServers', {})
for name, cfg in servers.items():
    stype = cfg.get('type', 'stdio')
    cmd = cfg.get('command', cfg.get('url', 'N/A'))
    print(f'  {name}: {stype} ({cmd})')
" 2>/dev/null || echo "Could not parse ~/.claude/mcp.json"

Verify each MCP server is live by calling a lightweight tool on each one. Run these checks in parallel:

Server	Health check tool	Call
ai-assistant	mcp__ai-assistant__time_now	No args needed
asana	mcp__claude_ai_Asana__asana_list_workspaces	No args needed
vercel	Use ToolSearch with +vercel to discover a read-only tool, then call it

For any claude.ai connectors (e.g., claude.ai Slack), use ListMcpResourcesTool with server set to the connector name. If resources are returned, mark it as live.

For each server, report:

• Live — tool returned a valid response
• Down — tool errored or timed out
• Auth expired — tool returned an auth/unauthorized error