Every DeepSeek Harness plugin

What it is — pi-dsh is a provider extension for Pi Coding Agent that registers DeepSeek Harness as a selectable provider, offering pool mode with in-process cordis tree and oneshot mode.

Who it is for — If you use Pi Coding Agent for coding tasks and want DSH to handle prompt, memory, and skill management, you can install pi-dsh. Users relying on HTTP LLM endpoints or not using Pi Coding Agent do not need it, as their needs and integration points differ.

Watch out — Installation requires Node.js >=22 and a working DSH install. Sandbox testing passes, registering successfully in a fresh profile. It executes shell commands but no other runtime issues were observed.

The verdict — I would install it because it integrates DSH in-process via pool mode directly into Pi Coding Agent without HTTP roundtrips.

What it is — This plugin adds agent_swarm tool and /swarm command for DeepSeek Harness, enabling Kimi-style parallel worker agents with coordinator synthesis.

Who it is for — When you use DeepSeek Harness to handle complex workflow orchestration that requires the model to explore and split multiple sub-tasks for parallel dispatch, it is suitable to install it. On the other hand, people who rely only on single-agent flows or do not intend to use swarm mode will find it unnecessary.

Watch out — Installation from source requires manually permitting the build script first. Sandbox testing passed with no obvious performance bottlenecks found. The static check notes that it executes shell commands, which is a normal functionality.

The verdict — I would install it because it lets the model handle sub-agent orchestration itself without manual parallel management.

What it is — Electron-wrapped desktop client for DeepSeek Harness that automatically starts the local service.

Who it is for — Users who run DeepSeek Harness Agent development in desktop environment without browser or terminal interaction should install it. People who prefer direct browser access to Harness Web interface do not need this desktop version.

Watch out — Sandbox test passed: installed in a new profile and registered by harness. From source installation need to manually allow the build script to run. No obvious issues found.

The verdict — I would install it because it can automatically start the local service and stop when closing the window, but only if I accept manually allowing the build script from source.

What it is — Provides a desktop client for DeepSeek Harness enabling mobile remote access.

Who it is for — If you need to remotely operate DeepSeek Harness from your phone or other devices. Users who run Harness agents directly on mobile without needing desktop backend should not install this client.

Watch out — Sandbox test passed in a new profile and registered into the profile. No configuration conflicts found after install. As a Beta version, plugins and persistent data may change, requiring backup before updates.

The verdict — I will install it because it enables remote control from mobile devices using outbound connections.

What it is — DSHDesktop packages the official DeepSeek Harness web UI into a Windows desktop application that runs without terminal installation while retaining native configurations and plugins.

Who it is for — Windows users who want to avoid terminal-based setup for DeepSeek Harness and prefer a desktop app with preserved native settings, sessions and plugins would benefit from this. Users on Linux or macOS do not need it as it is designed specifically for Windows desktop environments.

Watch out — Sandbox testing passed; the plugin registers successfully into the profile after installation. No obvious installation or runtime issues were found.

The verdict — I would install it because it provides a terminal-free Windows desktop version that integrates directly with existing DSH configurations and plugins.

What it is — DSH plugin providing GitHub OAuth device code login and repo operations.

Who it is for — Users who need to manage GitHub repositories in DSH can use this plugin for device code login. Users who do not require browser authorization can continue with manual token configuration.

Watch out — Installation requires copying files to the web profile directory and appending cordis.patch.yml registration. Sandbox test passed. No obvious issues found.

The verdict — I would install it because it enables GitHub access without manual token configuration, although it requires registering an OAuth app once.

What it is — This plugin adds Qichacha company data tools and a settings page QR login to DeepSeek Harness.

Who it is for — If you need enterprise information query tools in a DeepSeek Harness agent session, this plugin provides them. Users already obtaining enterprise data directly from official channels do not need it.

Watch out — Sandbox testing passed, installed in a fresh profile and registered by harness. No obvious issues found. Usage requires a personal Qichacha account and integral consumption.

The verdict — I would install it for its persistent mode after one-time setup, but only if you are prepared to install from source and manage the account.

What it is — It provides multi-agent orchestration for DeepSeek Harness by supplying role-bound agents with tools enforced via a runtime capability matrix.

Who it is for — If you are working on a project that requires multi-agent collaboration, using a specific model for planning and delegating tasks to expert agents, this plugin is suitable for you. If you prefer not to have enforced tool restrictions, this plugin does not match your needs.

Watch out — Sandbox test passed: installed in a fresh profile and successfully registered by harness. No obvious issues found.

The verdict — I wouldn't install it because it prohibits recursive delegation from sub-agents, which I find limiting for my needs.

What it is — This provides a self-contained desktop app for hosting the official web GUI of DeepSeek Harness while bundling its Node runtime and dsh server.

Who it is for — If you run DeepSeek Harness in a desktop environment and require a self-contained setup, this can be used. If you prefer to manage your runtime environment manually rather than using a bundled approach, you may not need this.

Watch out — Sandbox testing passed: it was installed in a new profile and registered by harness. The installation automatically executes scripts and runs shell commands. No obvious issues found.

The verdict — Given the sandbox test passed and registration in profile, I would install it, but I need to accept the auto script execution during source installation as a trade-off.

What it is — Seek Soul in Darkness is a desktop AI application based on DeepSeek Harness, seeking the soul of silicon life in darkness.

Who it is for — When you need to manage AI conversations using cross-session memory and Guardian state engine in your DSH desktop environment, Users who prefer pure code plugins rather than desktop applications shouldn't install it, because it requires a self-developed Electron shell and the full pre-made plugin package.

Watch out — Sandbox testing passed: it installed successfully in a new profile and was registered into the profile by Harness. No obvious pitfalls found.

The verdict — I would install it because it provides a self-developed shell and pre-made plugins, but only if you need a full desktop AI application on DSH.

What it is — Adds background customization to the DeepSeek Harness web GUI.

Who it is for — Users of the web profile in DeepSeek Harness should install it to customize the application background. Users of profiles that do not provide fs and webServer should not install it.

Watch out — Sandbox testing passed. Installation from source requires manually allowing its build script. No obvious issues found.

The verdict — I would install it because it provides persistent background customization in the web GUI.

What it is — This plugin provides vision capabilities to text-only DeepSeek models by describing images using an OpenAI-compatible vision model.

Who it is for — In the DeepSeek Harness web profile, when you need to use a text-only DeepSeek model to process image inputs and call tools, this plugin can provide vision capability complement. You do not need this plugin if you are using a non-web profile text-only model.

Watch out — Sandbox test shows the npm install command succeeds, but without bundle definition in package.json it is installed as an ordinary dependency and does not become a profile layer. Vision baseUrl, model and apiKeyEnv must be configured. No other obvious issues found.

The verdict — I would install it because it adds vision to text models, but it will not create an additional profile layer after installation and must be configured via the patch file as described in the README.

What it is — DeepSeek WebUI Studio is a visual workspace for building and previewing DSH WebUI plugins in isolation.

Who it is for — Developers who need to create or modify DSH WebUI plugins and want to test changes without affecting the stable host can use it. Users who do not develop DSH plugins do not need it.

Watch out — Sandbox testing shows it installs but blocks on dsh-harmony@0.1.3 build script, requiring manual allow-build flag. No other obvious issues found.

The verdict — I would not install it because the sandbox test requires an extra manual step to allow the build, and it has zero stars.

What it is — Provides a git commit/branch graph panel for the DSH Web UI sidebar.

Who it is for — When using the DSH Web UI sidebar for code review and needing to visualize git history of the current workspace, this plugin is suitable. Users not running the DSH Web UI do not need to install it.

Watch out — The plugin installs as a regular dependency rather than a profile layer. Manual configuration in cordis.patch.yml is required per the plugin's instructions. No obvious runtime issues were observed.

The verdict — I would install it for the git visualization feature in DSH Web UI code review, but only if you are using the web profile setup.

What it is — It integrates Fabric-style deterministic compaction, checked code execution, durable coordination, and live topology into DeepSeek Harness.

Who it is for — When handling durable coordination and live topology in DeepSeek Harness with need for structured summaries, installation is suitable. Users not involving Checked Code execution can opt for alternatives.

Watch out — Sandbox test shows harness did not register the plugin into profile after install. No obvious issues found.

The verdict — Sandbox test failed to register it into profile so I would not install directly. But if the registration issue is resolved, it would be worth checking again in durable coordination scenarios.

What it is — dsh-screenshot provides browser hotkeys for screenshots and an agent-facing capture tool for DeepSeek Harness.

Who it is for — When using DeepSeek Harness with multimodal models to analyze browser interfaces, this plugin supplies direct screenshot capabilities. Users without browser hotkey or agent tool needs can skip it.

Watch out — Sandbox test passed: installed into a new profile and registered by Harness. Requires adding the package to dependencies and dsh.profile.bundles in package.json, then restart the service. No obvious issues found.

The verdict — If your workflow involves real-time browser screenshots or agent tools, install it; otherwise not necessary.

What it is — Move Claude and ChatGPT history memories into DeepSeek Harness with verbatim evidence verification.

Who it is for — When you want to migrate memories accumulated from Claude or ChatGPT into DeepSeek Harness for recall_memory tool use. If your memory accumulation is only in local files, you do not need this plugin.

Watch out — Tested successfully in a new profile where the plugin was registered into harness. The cost for one move is about ¥ 2.5 to ¥ 5 depending on the model service provider. Claude memories.json import costs zero tokens.

The verdict — I would install it because it moves Claude and ChatGPT memories into DSH for model recall and supports exporting to other memory plugins.

What it is — This is a TypeScript plugin, a fork of deepseek-harness-desktop for DeepSeek Harness, kept in sync with upstream via the git remote.

Who it is for — When you need to build the desktop application of DeepSeek Harness from source, you can install this plugin. Users who do not use the desktop application do not need to install it, because they can run it via npm.

Watch out — Installation will automatically execute scripts. The sandbox test passed, it was installed in a new profile and registered into the profile by harness.

The verdict — I would install it because the sandbox test passed.

What it is — dsh-theme supplies a theme pack with 30 ready-to-use CSS themes for DeepSeek Harness Web UI, including light and dark variants, integrated via the settings → general slot.

Who it is for — When working in the DeepSeek Harness Web interface and needing to switch visual themes to fit personal preference, this plugin can be added directly. If your workflow is limited to pure text input without any visual customization needs, then this plugin is not required.

Watch out — Sandbox test confirmed successful registration into harness and presentation in the general settings slot after installation in a fresh profile. Static check noted shell command execution but no obvious issues found. Installing from source requires manually allowing the build script to run.

The verdict — I would install it because it offers a range of theme options and sandbox testing confirmed successful registration in a fresh profile, provided the build script can be manually allowed to run from source.

What it is — dsh-login provides a single-password authentication gateway for the DeepSeek Harness Web GUI.

Who it is for — When the DSH web service is exposed on 0.0.0.0 or a public network and you need to add password protection for Web GUI access, you can install this plugin. If you run DSH using a non-web profile, you do not need to install this plugin.

Watch out — Sandbox test passed: the plugin was installed in a new profile and registered into the profile. It will execute shell commands and requires providing an API key or token, but no obvious pitfalls were found.

The verdict — I would install it if I need to protect the web GUI on a publicly exposed DSH, because it takes over the fallback handler and re-provides the webRuntime service.

What it is — DSH Inline Comments lets you add numbered comments next to selected text in DeepSeek Harness assistant replies and submit them together via the official input box.

Who it is for — When reviewing long DeepSeek Harness assistant replies and needing to make comments at multiple spots, install this plugin. Skip it if you are not willing to manually allow the GitHub build script.

Watch out — It installs successfully in the browser but the build script is blocked by pnpm, requiring manual allowance to complete registration. It requires DeepSeek Harness 0.1.0-rc.6 or higher. No other obvious issues were observed.

The verdict — I would install it because it routes selected comments straight into the official input box, but only after manually permitting the build script; otherwise the full workflow cannot be used.

What it is — PPTKit Presentation skill converts topics, documents, spreadsheets, images, and presentations into editable PowerPoint decks with local browser preview and explicit PPTX export.

Who it is for — In a DeepSeek Harness agent, when you need to convert a quarterly report PDF and Excel metrics into an executive presentation, it is suitable to install. If your workflow does not rely on local-first browser preview, you can skip it.

Watch out — Sandbox test passed: after installing into a fresh profile, it can be registered by Harness. Requires Node.js 20 or higher, pnpm, and dsh CLI. Static check notes it will execute shell commands; source install requires manually allowing the build script.

The verdict — I would install it because it allows reviewing slides in the browser before explicitly generating an editable PPTX, which helps control presentation quality, but the prerequisite is having dsh CLI and a Node.js environment set up.