ZekaiShi

#1·evo-subagent

ZekaiShi

5

Unified DeepSeek Harness plugin: role-based subagent routing + per-agent evolution (prefercmd/memory as knowledge allow/deny lists), so repeated tasks start from proven commands and save tokens. Unified subagent routing and evolution: prefercmd/memory serve as knowledge allow/deny lists, saving tokens.

#2·smart-subagent

ZekaiShi

—

Maps a stable agent_key to an exact provider/model pair declared in role Markdown files, with per-workspace evolution of prefercmd and memory files.