Hacker News Reader

Subagent personalities, skills, structured output generation, customizable context cleanup, etc.

Most of those seem kind of useless... I probably do not know how to use them correctly, so, what would you recommend by way of subagents and skills?