dsh-plugin-mlquant-benchmark

Running and reporting a quant benchmark inside DeepSeek Harness used to mean juggling fixed seeds, panel sizes, thread counts, JSON field checks, and a rigid issue template by hand. dsh-plugin-mlquant-benchmark collapses that whole workflow into four DSH tools that operate on the ml-quant-trading protocol v1 CPU benchmark: execute the fixed-parameter run and emit benchmark-v1.json, render the JSON artifact as a compact Markdown table, validate every required protocol field and flag variance warnings, and draft an issue-ready report body from the artifact without ever posting to GitHub. The plugin is deliberately narrow—it adds no trading agent, calls no market-data API, configures no model provider, and makes no backtest or performance claims—so the throughput numbers stay evidence, not alpha. It is built for quant and agent engineers who want a reproducible, end-to-end benchmark-to-report path inside DSH without pulling in heavyweight infrastructure.