DeepSeek Harness Plugin

CZ-ZL/duo#dsh-plugin

Stars ★ 0 Category Development & Runtime Added 2026-09-18

DSH plugin for bounded agent evaluation: compare an original with candidate prompt or supported configuration changes against user-supplied checks, record evidence and costs, and leave adoption to the user.

Install

# from GitHub (first run asks for allowBuilds approval — follow the hint, retry)

dsh plugin --profile web add github:CZ-ZL/duo#path:/dsh-plugin

Any plugin you install runs third-party code with your own permissions — it can read your files, use your credentials, and reach the network, and tool approvals don’t sandbox it. GitHub-sourced plugins also run build scripts at install time — pnpm blocks those until you allow them, so an install can stop with ERR_PNPM_GIT_DEP_PREPARE_NOT_ALLOWED or ERR_PNPM_IGNORED_BUILDS; dsh prints the exact key to add under allowBuilds in your profile’s pnpm-workspace.yaml, and the install works on the next run. Allowing a build is a trust decision: only install sources you trust, and pin a commit (github:owner/repo#sha).

README

English · 简体中文

DUO lets a DSH Agent evaluate an original prompt or supported component, try candidate changes against supplied tests, and return the changes, decisions and costs.

Start here: install → free check or real task → saved report

The Quickstart includes every command, the supplied task/evaluator and the owner-provided model configuration. An independent Caller has completed the real starter through this public guide; the original and candidate tied, so the original was retained. The local path checks product operation without model calls.

Developer preview: Linux / Node24+, tested DSH0.1.2-rc.1 / Cordis4.0.2. Version 0.6.3 is an unreleased source preview; the Quickstart packages its public source into an installable tarball. Previous released builds have versioned GitHub assets. Do not install the repository root or assume an npm release. Installation may fetch dependencies; it does not configure a model account or authorize experiments.

The default bundle offers preparation until compatible work providers are attached. The shipped starter supplies its own composition. Built-in model providers support persona/prompt; other Targets require matching providers. No candidate is automatically adopted. DUO inner costs exclude Calling Agent inference and local resources. Python/YAML compatibility remains behind /legacy; native use needs no Python.

Content from the project README on GitHub ↗

Links

More in this category

View the whole category →

Community comments

Comments are public GitHub Discussions. Loading them connects to GitHub and Giscus; a GitHub account is required to post.