feat: add ai-to-ai and emotional-intelligence skills

[AdelElo13]

Summary

Two new community skills:

1. ai-to-ai (567 lines)

HTTP-based inter-agent communication protocol. Teaches Claude how to send tasks to and receive tasks from other AI agents.

• Complete protocol spec (POST /ai-to-ai with JSON task field)
• Working server examples (Python Flask + Node Express)
• Client examples (curl, Python requests, async)
• Bearer token security with RBAC
• Multi-agent patterns: fan-out, pipeline, supervisor
• Google A2A protocol (JSON-RPC 2.0) integration
• Error handling and timeout patterns

2. emotional-intelligence (195 lines)

Frustration detection and tone adaptation. Makes Claude a better conversational partner.

• 3-layer detection: keyword signals (regex), behavioral signals (message patterns), contextual signals (conversation history)
• Tone adaptation rules per frustration level (mild → moderate → high)
• Energy matching (match user's message length/tone)
• Conversation arc tracking (escalation, de-escalation, recovery)
• Banned phrases with alternatives ("I understand your frustration" → just fix it)
• 5 good vs bad response examples
• Cultural awareness (Dutch directness, English indirectness)
• De-escalation principles

Both inspired by the Jarvis AI operating system and the DockWright macOS agent.

Test plan

• ai-to-ai: Server examples tested with curl and Python client
• emotional-intelligence: Detection patterns tested against real frustrated user messages (Dutch + English)