dsh-benchmark: Generating Reproducible Benchmark Evidence for DSH Tools and Plugins

The dsh-benchmark is a benchmarking plugin for tools and plugins targeting the DeepSeek Harness (DSH), designed to provide reproducible and deterministic benchmark evidence. Centered around fixed use cases, it leverages three core commands—`inspect`, `run`, and `compare`—to enable protocol metadata inspection, fixed use case execution, and baseline regression comparison, while generating content-addressed reports. After installation, it registers corresponding tools within DSH and supports interaction via both the CLI and a standalone stdio MCP server interface. The workflow is as follows: first, inspect the manifest and target fingerprint; then, run fixed use cases to produce artifacts; finally, compare the current report against the baseline to evaluate performance or behavioral stability. Exit codes carry clear meanings, respectively indicating pass, scoring failure, or operational errors. This plugin is applicable