fix(templates): improve AGENTS.md template formatting and accuracy

[edenreich]

Summary

This PR improves the AGENTS.md template to fix inaccuracies and enhance readability:

• Removed irrelevant curl examples - The template previously showed fictional skill execution endpoints that don't exist in the A2A protocol
• Fixed API endpoints - Now correctly shows only POST /a2a for JSON-RPC instead of individual skill endpoints
• Corrected environment variables - Updated to use ADK's actual AGENT_CLIENT_* format instead of provider-specific API keys
• Port consistency - Now uses the configured port from the ADL spec throughout the document
• Improved formatting - Aligned all comments in the project structure tree and removed excessive blank lines

Changes

Before

# Execute create_dashboard skill
curl -X POST http://localhost:8080/skills/create_dashboard \
  -H "Content-Type: application/json" \
  -d '{"input": "your_input_here"}'

Environment variables:

- `DEEPSEEK_API_KEY` - API key for deepseek
- `PORT` - Server port (default: 8080)

After

Refer to the main README.md for specific skill execution examples and input schemas.

Environment variables:

- `AGENT_CLIENT_PROVIDER` - LLM provider (deepseek)
- `AGENT_CLIENT_MODEL` - Model name (deepseek-chat)
- `AGENT_CLIENT_API_KEY` - API key for your LLM provider
- `PORT` - Server port (configured: 8443)

Test plan

• Generated a test project with --ai flag
• Verified AGENTS.md has correct API endpoints
• Verified environment variables match ADK documentation
• Verified all comments in project structure are aligned
• Verified port numbers are consistent throughout

🤖 Generated with Claude Code

[ig-semantic-release-bot]

🎉 This PR is included in version 0.24.1 🎉

The release is available on:

• v0.24.1
• GitHub release

Your semantic-release bot 📦🚀