Every DeepSeek Harness plugin

What it is — Provides bounded search workflow for DeepSeek Harness agent, making source scope and retained evidence visible.

Who it is for — When using a search-capable agent for factual queries, this plugin makes source scope and evidence visible. If you do not use a search-capable preset or do not care about evidence visibility, you may not need it.

Watch out — Sandbox testing passed with successful installation and registration in a fresh profile. No obvious issues found. Requires DEEPSEEK_API_KEY and a search-capable preset.

The verdict — I would install it because it can display unresolved gaps in search evidence, but only if you need to verify source scope.

What it is — dsh-shell-termux registers as ctx.shell on Android to provide Termux bash execution capability for DeepSeek Harness.

Who it is for — In the scenario of executing bash tools with DeepSeek Harness on Android, this plugin is suitable when Termux support is required. If you run DeepSeek Harness on non-Android platforms without needing Termux environment injection, you may skip this plugin.

Watch out — Installation requires placing the plugin in the profile's node_modules and modifying cordis.patch.yml to disable the bash-sandbox provider. Sandbox testing passed: it was installed in a fresh profile and registered by harness. No obvious pitfalls were found, but bashPath must point to the Termux binary.

The verdict — If running DeepSeek Harness on an Android device for shell tasks, I would install this plugin because it offers probe diagnostics and explicit environment injection.

What it is — DSH plugin verification CLI runs mock-llm cycle to check agent behavior.

Who it is for — Plugin authors use this tool to verify their plugins' runtime behavior when adding the Verified badge. Ordinary users who only want to consume verified plugins will not need this tool.

Watch out — Sandbox test shows it installs as ordinary dependency without becoming a profile layer. No obvious pitfalls found.

The verdict — If I were a plugin author, I would install it because it serves plugin authors' verification needs, but I would use it as described in the README without extending the profile layer.

What it is — Injects legacy-browser polyfills into pages via webserver hook to fix workspace directory picker RPC cancellation and other RPC flows on older kernels.

Who it is for — When using MuMu bundled browser or older WebView for workspace directory picker it's suitable to install. Users on modern browsers supporting AbortSignal.any() or using native file pickers can choose not to install.

Watch out — Sandbox test passed in new profile, plugin registered by harness. Static check indicates execution of shell commands. No other obvious issues found.

The verdict — I would install it if using unsupported old WebView for RPC operations, because it provides the needed polyfills, but not worth it on a supported kernel.

What it is — Integrates OpenCode Go LLM provider into DeepSeek Harness profile, adding model routes, settings cards, and catalog updates.

Who it is for — When using OpenCode Go models for inference in the DeepSeek Harness web profile, this plugin provides Web Connect cards and stale-while-revalidate catalog updates. If your profile already includes other LLM provider routes, no need to manage this Git dependency.

Watch out — After installation, the harness does not register this plugin into the profile. No obvious installation or runtime issues were found.

The verdict — I would not install it now because it does not register after testing. I would consider it again if the profile update supports loading such bundles.

What it is — Import, validate, manage, and migrate Agent Skills to DeepSeek Harness.

Who it is for — When using DeepSeek Harness Web UI for AI coding, users who need to import Agent Skills from Claude Code, Codex, and other agents are suitable to install it. If you do not need to migrate skills from other tools to DeepSeek Harness, you can continue operating in the original tool.

Watch out — Successfully tested in a new profile where the harness registered it. No obvious pitfalls found.

The verdict — I would install it because it supports batch migration and direct management of skills in the DSH Web UI, but only if the skills need to be migrated from other agents.

What it is — This is a DeepSeek Harness Web approval enhancement plugin that supports rejecting with comments and terminating the round to let the model rethink with the comments.

Who it is for — It is suitable when using DeepSeek Harness in Web mode to handle approval processes. You do not need to install it if you are not using the Web mode for approval or in non-web DSH scenarios.

Watch out — Sandbox test passed in a new profile and was registered by harness into profile. Static checks found no defects. No obvious pits detected.

The verdict — I would install it because it is pure JavaScript with zero third-party dependencies, but the premise is that it must be installed in the web profile and restarted afterward.

What it is — This is a DeepSeek Harness plugin that integrates Kiro account to serve Claude series models without separate API key.

Who it is for — When using DeepSeek Harness for Claude model driven development tasks, this plugin allows direct integration via Kiro account without managing separate API key. For users not using Claude series models, this plugin's integration capability is not applicable.

Watch out — No obvious pitfalls found in sandbox test; it passed when installed in a new profile and registered with harness. Source install requires manually allowing its build script. Claude model usage requires configuring proxyUrl to permit network egress.

The verdict — I would install it because it can directly serve models on the same machine.

What it is — Add a one-click button to the left of the model selector in the input box to optimize fuzzy requirements into executable prompts for AI agents.

Who it is for — This plugin suits users who input vague requirements for software development or design tasks using an agent with planning and execution capabilities in DeepSeek Harness. Users who do not rely on agent planning and execution for their tasks do not need it.

Watch out — Installation from source requires manually allowing the build script. Sandbox test passed: installed in a new profile and registered in harness. The plugin will execute shell commands.

The verdict — I would install it because it converts vague requirements into structured executable prompts that improve agent task completion.

What it is — dsh-zen provides Zen mode for DeepSeek Harness Web UI, enabling toggle via Ctrl+Shift+Z to hide sidebar and topbar.

Who it is for — When using DeepSeek Harness agent to handle complex tasks and needing focus control, this plugin is suitable to install. If you prefer to keep the full interface with multiple panels visible, you may not need to install it because it hides the sidebar and topbar.

Watch out — Sandbox testing passed. It was installed in a new profile and registered by harness. It executes shell commands but no obvious issues found.

The verdict — I would install it because the sandbox test passed and it was registered in a new profile, but only if I manually allow the build script when installing from source.

What it is — Integrates DSH update check and upgrade functionality into the Web UI settings page, allowing one-click operations.

Who it is for — Users managing DSH plugins via the Web UI settings page are suitable for installing this plugin to check for updates. Users who update DSH versions directly via npm do not need this plugin because they do not rely on Harness's Web UI integration.

Watch out — Sandbox test passed: installed in a new profile and registered into harness profile. No obvious issues found.

The verdict — I would install it because it provides a direct Web UI interface for checking and upgrading DSH versions.

What it is — DeepSeek Harness plugin allowing use of OpenAI ChatGPT subscription without API key.

Who it is for — Suitable for users who want to use ChatGPT models in DeepSeek Harness without API key setup. Those who already have API tokens for other providers don't need it as they can use direct configurations.

Watch out — Sandbox test passed with successful installation in a new profile and registration by harness. Static check indicates need to provide API key or token. No other issues found.

The verdict — I would install it because it reuses the local Codex CLI auth.json for login, assuming you have an OpenAI subscription.

What it is — Displays DeepSeek API account balance in DeepSeek Harness with manual refresh and 10-minute cache.

Who it is for — If you are using DeepSeek Harness web profile and need to show DeepSeek API account balance, install it. If you prefer to manage API keys directly in the browser without plugin integration, do not install it because the key will not be returned to the browser.

Watch out — Sandbox test passed in fresh profile where the plugin was registered. No obvious issues found.

The verdict — I would install it because the sandbox test passed in a fresh profile and it maintains security.

What it is — Provides workflow skills for AI coding agents, including state machine, risk grading, and code graph impact analysis.

Who it is for — When handling feature change requests in an existing project using DeepSeek Harness agent, these skills are suitable for providing safe incremental modifications. For users working on new project initialization or small personal projects, these skills are not suitable because they focus on engineering control.

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

The verdict — I would install it because it provides cross-session plan state and vector knowledge capture, which adds value for long-term projects.

What it is — Floating widget plugin displaying real-time DeepSeek API balance and service status.

Who it is for — When developing in DeepSeek Harness, you need to monitor DeepSeek API balance and status in real time. If you primarily use other sidebar tools, you don't need this plugin.

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

The verdict — I would install it because it provides real-time monitoring without exposing the API key.

What it is — Practice DSH plugin package, covering plugin lifecycle, tools, services, events and waterfall middleware concepts.

Who it is for — Developers who want to practice tool definition and service injection in DSH plugin development scenarios are suitable for installing it. Those who do not want to learn plugin development do not need to install it.

Watch out — Sandbox test shows that from-source installation requires manually allowing the build script to complete registration. No other obvious issues found.

The verdict — I would install it because it offers systematic examples, but I must manually allow the build script first.

What it is — Provides secure remote access to the DeepSeek Harness Web GUI using password-protected Cloudflare Tunnel.

Who it is for — When you need to securely access the DeepSeek Harness Web GUI over the internet with a password. Users who access it only within the local network do not need this plugin.

Watch out — From source installation, you must manually allow the build script. Set up the shared password afterward for tunnel access. No other issues detected.

The verdict — I would install it because it enables secure remote GUI access, but I must manually permit the build script first.

What it is — It is a DeepSeek Harness skill that provides an orchestration engine for long tasks with cheap models, using disk state machine and mechanical gates for reliability.

Who it is for — When you need to use cheap models for long chain tasks and require structured reliability, it is suitable. Users who only need short or single-step tasks do not need it.

Watch out — Static check indicates it will execute shell commands. From GitHub source installation you need to manually allow the build script. Sandbox test passes in a new profile and registers successfully. No obvious pitfalls found.

The verdict — I would install it because it allows cheap models to reliably handle long tasks through structure rather than model strength. The premise is that your task requires multi-step orchestration.

What it is — Exposes DeepSeek Harness host runtime status through HTTP endpoints and UI badge.

Who it is for — In environments running DeepSeek Harness for web applications that require viewing host process status and memory utilization, it's suitable to install. Users who do not want to integrate status alerts into the browser UI can choose not to install.

Watch out — Sandbox test passed: successfully installed and registered in a fresh profile. It will execute shell commands, no obvious issues found.

The verdict — I would install it because it provides real-time push of host status alerts.

What it is — DeepSeek Harness plugin that simulates vision for text-only LLMs via ascii-art from images.

Who it is for — Users performing image analysis with text-only LLMs in web profiles can use it. People needing true vision models or native multimodal capabilities do not need it.

Watch out — Installation takes one npm command. Sandbox testing succeeded with registration in a new profile. No obvious issues found.

The verdict — If you use text-only LLMs and need occasional image analysis, install it as a workaround; otherwise it is not worth it because it is not real vision.

What it is — DeepSeek Harness vision plugin equips text-only models with vision using the vision_analyze tool and automatic image transcription.

Who it is for — When using DeepSeek Harness to switch to text-only models such as deepseek-v4 to analyze uploaded images. Users who mainly use native vision models can choose not to install this plugin.

Watch out — Sandbox test passed: the plugin was installed in a new profile and registered into the profile by harness. No obvious issues found.

The verdict — I would install it because it enables vision for text-only models in DeepSeek Harness but the premise is that the vision model is correctly configured.

What it is — Integrates a workspace file browser into DeepSeek Harness, supporting tree directory browsing, search, and online editing.

Who it is for — When using DeepSeek Harness for project development and needing to browse and manage files in the layout panel, this plugin is suitable to install. If you prefer local file management tools, you can skip installing this plugin because it requires browser-side integration into the layout area.

Watch out — Sandbox testing passed. It registers in a new profile. Installation from GitHub source involves executing shell commands.

The verdict — I would install it because it supports direct file editing in conversation tabs.

What it is — This DSH plugin lets you adjust LLM provider retry count via chat command.

Who it is for — When you use dsh-llm-pi-ai provider and want to adjust retry count via chat. If you use native DeepSeek adapter (dsh-llm-deepseek), because this plugin does not support it yet.

Watch out — Sandbox test succeeded with registration in a new profile. No obvious issues found. However, unlimited retries can hammer a provider that is down.

The verdict — I would not install it because it is limited to the pi-ai adapter and unlimited retries risk hammering failing providers indefinitely.

What it is — DSH Web three-column layout plugin that purifies dialogue flow by hiding tool call cards while retaining thinking process display, plus left workspace file tree with git status coloring.

Who it is for — When you are performing software development tasks with DeepSeek Harness and need left workspace file tree support for git status tracking, this plugin is suitable to install. For users primarily engaged in pure reasoning conversations, the official default layout can be chosen instead because this plugin adds extra left panel enhancement.

Watch out — Sandbox testing passed: successfully installed in a new profile and registered to harness. Compatible with @deepseek-ai/dsh 0.1.0-rc.6 series.

The verdict — I would install it because it renders a bottom status bar via portal that displays selected model balance instead of official turn/step counter.