@SnowRikka

DeepSeek Harness plugin: run subagents on a local llama.cpp model via the OpenAI Responses (/v1/responses) protocol — LLM adapter + delegation skill