DeepSeek Harness Research Lab Plugin (dsh-research-lab)

`harzva/dsh-research-lab` is a standalone workflow plugin for DeepSeek Harness (DSH) that aims to address the difficulty of pinning down and reproducing experimental configurations. It provides a Manifest-first control plane that supports managing multiple inference runtimes (such as ReAct and Plan Guided) within the Web interface, and allows users to pin baseline parameters, generate ablation matrices, and output deterministic JSON configurations with checksums. The plugin does not invoke models directly; it is designed to help researchers rigorously compare and reproduce different reasoning strategies.