Add skill-audit-mcp to Security

[eltociear]

Adds skill-audit-mcp under Security.

Scans MCP servers, AI agent skill files, and plugins for 68 attack patterns across 4 severity levels:

• CRITICAL — credential exfiltration, seed-phrase harvest, download-and-execute
• HIGH — arbitrary code execution, auth bypass, identity impersonation
• MEDIUM — prompt injection, obfuscation, privilege escalation
• LOW — external URL refs, broad filesystem access

Output: SARIF 2.1.0 → GitHub Code Scanning Security tab.

Usage:

- uses: eltociear/skill-audit-mcp@v1
  with:
    path: '.'
    severity: 'MEDIUM'
    fail-on: 'HIGH'

Composite action, no Docker/Node setup overhead. Released v1.0.1 on 2026-05-11.