Every DeepSeek Harness plugin

What it is — Adds Chat/Work switch and managed Chat workspace to DeepSeek Harness web GUI, allowing conversations without project directory selection.

Who it is for — When you want to chat in the DeepSeek Harness web GUI without choosing a project directory. If your tasks are always tied to specific project directories, this plugin's chat mode won't be utilized.

Watch out — Sandbox test shows the build script is blocked by pnpm requiring manual approval to install from source. No other issues found.

The verdict — I would install it because it enables chat sessions without a project workspace in DeepSeek Harness, although installation from source requires manually allowing the build script.

What it is — Adds inline controls to DSH model directory editor for configuring image support and reasoning efforts per model.

Who it is for — Users who maintain custom model lists in DeepSeek Harness and need to adjust image support or reasoning efforts per model find it suitable. Users who only operate within the official llm-deepseek namespace do not need it.

Watch out — Sandbox test shows installation can complete but pnpm blocks the build script from github:ztlovelsw/dsh-model-profile, requiring manual approval before full registration. No other issues detected.

The verdict — I would install it because it enables inline configuration of image support and reasoning efforts for custom models directly in the editor, avoiding manual settings.yaml edits, but only after manually allowing the build script.

What it is — Drag files into the DSH window to write real disk paths into the input box

Who it is for — Users who need to insert file or folder paths into the DSH WebUI session input box are suitable for installing this plugin. Users who use desktop applications rather than browser drag and drop can complete file path handling through system native file selection

Watch out — Sandbox test passed, installed in fresh profile and registered into the profile. No obvious pitfalls found.

The verdict — I would install it because it supports directly reading real complete paths from the Windows drag and drop system

What it is — The plugin retrieves evidence from local user directory storage to answer DSH knowledge questions and supports switching to strict mode refusing insufficient evidence.

Who it is for — Users who want to explain knowledge in DSH chats using the current model and prioritize local evidence retrieval are suitable for installing it. Because the plugin does not require any database, vector model, account login, or project token and does not connect to any external projects, people who need external knowledge databases or cloud services do not need it.

Watch out — Sandbox test passed, successfully installed in a new profile and registered by Harness. No obvious issues found.

The verdict — I would install it because it requires evidence to support answers and runs entirely locally without needing any external configuration.

What it is — DSH Provider Hub is a native plugin that provides official account OAuth, API channels, model discovery, failover, and logs.

Who it is for — When using DeepSeek Harness to proxy multiple API providers and need failover plus unified logging, the plugin is suitable. It can be skipped when a single provider is stable and no logging is required.

Watch out — Installation is from GitHub source, and the first start generates a client access key that must be saved. Sandbox testing succeeds in a new profile and registers successfully. No other issues found.

The verdict — I will install it because it manages official accounts and multiple API channels with automatic failover, which directly helps multi-channel usage scenarios.

What it is — dsh-wechat-mp-studio is a DeepSeek Harness plugin for WeChat Official Account content production, offering structure signature rotation and gpt-image image pipelines.

Who it is for — If you are producing recurring WeChat Official Account content such as morning greetings, goodnight wishes, solar-term cards, homilies and blessing cards using DeepSeek Harness, but worry about platform low creativity flags, this plugin is suitable.

If you are not involved in any WeChat Official Account content creation, this plugin is irrelevant.

Watch out — It requires installation from GitHub source. Sandbox test succeeded: installed in a fresh profile and registered into the Harness profile. No obvious pitfalls found.

The verdict — I would install it because it provides an executable anti-homogenization production pipeline.

What it is — This is a zero-dependency DeepSeek Harness plugin for Windows that automatically shuts down after tasks finish, with a cancellable countdown button.

Who it is for — If you're running long tasks such as batch migration or model generation using DeepSeek Harness and want automatic shutdown on completion, this plugin is suitable. Users without Windows systems don't need it.

Watch out — No obvious issues were found in testing. It successfully registered in a fresh profile under sandbox conditions. Options allow customization of the countdown period and margin.

The verdict — I wouldn't install it because it's Windows-specific and requires re-enabling after each restart of DeepSeek Harness.

What it is — Adds 3D lever next to DSH send button to trigger pixel gacha.

Who it is for — Users chatting with model in DeepSeek Harness who want to observe reasoning signals for fun fit. Not suitable for users requiring high accuracy in model outputs, because it settles only on text signals.

Watch out — Sandbox test passed; plugin registered successfully in new profile. No obvious issues found.

The verdict — I wouldn't install it, as it does not save the full reasoning chain, only for entertainment.

What it is — This plugin integrates a token usage statistics panel into the DeepSeek Harness settings sidebar.

Who it is for — Users who need to track token consumption, balances, and cost estimation across multiple LLM providers such as DeepSeek official, OpenAI, and Anthropic in DeepSeek Harness would find this plugin suitable. Users who only use a single LLM provider and do not require cross-channel statistics can skip installing this plugin.

Watch out — Sandbox testing confirmed successful installation in a new profile and registration by harness. Static checks indicate it will execute shell commands. Usage queries for OpenAI and Anthropic require organization-level API keys.

The verdict — I would install it because it provides comprehensive token statistics, balance queries, and cost estimation across channels, which is useful for multi-provider scenarios, although some channels require organization-level API keys.

What it is — This plugin supports selecting text in AI replies in the DeepSeek Harness web GUI and referencing it to the next message.

Who it is for — Users who need to reference specific content from AI replies when using the DeepSeek Harness web GUI can use this plugin. Those who do not use the DeepSeek Harness web GUI do not need to install this plugin.

Watch out — Sandbox testing shows the build script is blocked by pnpm, preventing harness registration. Manual permission for github:skyhancloud/dsh-client-ui-quote is required to complete installation.

The verdict — I would install it because it allows clean quoting without adding text to the draft, but the build script requires manual approval before installation.

What it is — dsh-invest-research is a plugin for DeepSeek Harness that delivers full-chain A-share investment research, including real-time quotes, Buffett-style value scoring, risk scanning, and institutional consensus.

Who it is for — If you are using DeepSeek Harness model tools to analyze A-share stocks such as 600519 and call invest_research to obtain structured research reports, you can install this plugin.

If your research only involves Hong Kong or US stocks or does not require multi-layer data integration, you do not need to install it.

Watch out — Installation from GitHub source requires manually allowing the build script. Sandbox testing confirmed successful registration in a fresh profile, and no obvious issues were found.

The verdict — I will install it because it provides independent fault tolerance across data layers for A-share research.

What it is — DeepSeek Harness plugin adds Workspace Only switch, requiring user confirmation for file access outside workspace in any mode.

Who it is for — A user working in DeepSeek Harness workspace-write mode who is concerned about external file access is suitable to install this plugin. If a user only operates within the workspace and does not need additional security layers, this plugin is unnecessary.

Watch out — Sandbox testing passed. It was installed in a fresh profile and registered into the profile by Harness. The plugin executes shell commands during installation.

The verdict — If I were you, I would install this plugin because it forces user confirmation in any mode by bypassing approval strategies.

What it is — This plugin enables publication of workspace files into chat artifacts with inline image previews and downloadable cards.

Who it is for — When using DeepSeek Harness to create user-facing files, this plugin enables inline previews and download cards. Users who do not use the chat interface for artifact display do not need to install this plugin.

Watch out — Sandbox test passed. The plugin was registered into the profile after installation in a fresh profile. No other costs were observed.

The verdict — I would install it because it keeps artifacts visible after session reload.

What it is — This is a voice plugin for DeepSeek Harness. It provides per-message voice playback, a session-level auto-announce toggle, and dictation via microphone. It supports cloud TTS and ASR from Alibaba DashScope or Volcengine, which can be freely combined, with a fallback to the browser's Web Speech API.

Who it is for — Developers working locally with DeepSeek Harness for code debugging who need voice feedback and input assistance can install this plugin. Users who rely exclusively on text-based interactions do not need voice capabilities.

Watch out — It installs and registers successfully in a fresh profile. Environment variables SPEECH_DASHSCOPE_API_KEY or SPEECH_VOLCENGINE_API_KEY must be set. Building from source requires executing the build script first. No additional issues were observed in testing.

The verdict — I would install it for local development workflows because it allows independent selection of TTS and ASR engines along with system fallback.

What it is — dsh-env-profile plugin automatically detects machine tools, paths, network and disk, stores cross-session archive and injects it into system prompt with fixed budget.

Who it is for — If you do local development or operations using DeepSeek model and need environment context about available tools, paths and network delay, this plugin fits.
If your sessions only use cloud APIs with no local tool dependency, you can skip it.

Watch out — Sandbox test passed: installed in fresh profile and registered by harness, injected environment summary into prompt. No obvious issues found.
Installation requires manually allowing build script and physical copy to replace pnpm junction; npm or git source install avoids this.
Known limits: Windows-only detection, knowledge entries require manual entry, single-process write needs care.

The verdict — I would install it because it stabilizes environment injection and reduces token use, but only when my work depends on local tools and paths.

What it is — dsh-engineering-workflow provides an engineering workflow layer for DeepSeek Harness, with five gated phases.

Who it is for — If you are using DeepSeek Harness for engineering tasks that require TDD and parallel subagent execution, you can install this plugin. If you are using other frameworks that do not rely on DeepSeek Harness delegation tools, you can skip it because it depends on DeepSeek Harness workflow orchestration.

Watch out — After installation, dsh needs to be restarted to mount the host plugin. Sandbox testing passed in a fresh profile with successful registration. It will execute shell commands.

The verdict — I would install it because it integrates workflow skills with dsh's preset system.

What it is — DeepSeek Harness plugin replaces agent retry with falsification-driven correction loop to record auditable evidence.

Who it is for — Developers using DeepSeek Harness for agent sessions need to replace agent retry with falsification loop to record auditable evidence. If not running DeepSeek Harness, the plugin has nothing to attach to.

Watch out — Sandbox test shows that after installation, harness did not register the plugin into profile. No other obvious pits were found.

The verdict — If I were you, I would not install it because the sandbox test shows it was not registered into profile; but if cordis.yml has the plugin declaration, it can turn agent retry into auditable evidence.

What it is — In DeepSeek Harness web chat, it inserts an inline notification before answering larger questions to estimate token usage and DeepSeek API costs, then updates the same line to actual bill after.

Who it is for — It suits users handling large questions in DeepSeek Harness web chat with v4-flash or v4-pro models. It does not suit users who prefer a clean chat interface without any insertions because the plugin inserts estimates at its thresholds.

Watch out — Sandbox test passed. It was installed in a fresh profile and registered into the profile. No obvious pitfalls found.

The verdict — I would install it because the session-based calibration makes output token estimates more accurate the more it is used.

What it is — The plugin enables manual selection of input types for custom models in DeepSeek Harness.

Who it is for — When creating a custom model in DeepSeek Harness that supports image input, input capability declaration is missing. For pure text model users, from another angle, if capabilities are default decided by the provider, then this plugin is not needed.

Watch out — Sandbox test shows it can be installed but the build script is blocked by pnpm, requiring manual approval to finish registration. Static checks indicate shell command execution. No obvious issues found.

The verdict — I would install it from source because it allows specifying input types for models, but I must manually allow the build script first.

What it is — Real-time display of token usage and balance for DeepSeek Harness sessions.

Who it is for — In users who perform dialogue in the DeepSeek Harness web interface, this is suitable to install. In users who do not perform model calls, this is not suitable because it relies on official token reporting.

Watch out — When installing from source, manually allow the build script. Tested successfully in new profile. No obvious issues found.

The verdict — I would install it because it provides real-time balance refresh.

What it is — Provides a turn-based conversation timeline navigator for the DeepSeek Harness Web UI with bookmarks, history loading, keyboard controls, and mobile support.

Who it is for — Users handling long conversations on DeepSeek Harness Web UI who need quick history navigation are suitable for installing this plugin. Users on older DSH Web client versions can skip installing it.

Watch out — Installation requires the dsh plugin command or a PowerShell script, possibly involving page refresh and restarting the DSH Web process. Sandbox test passed in a new profile. No obvious pitfalls found.

The verdict — I would install it because it supports automatic history loading and language switching, but the premise is using a supported DSH version.

What it is — This DeepSeek Harness plugin notifies AgentTeams agents of task completion by having GINKA make WeChat-style voice calls to Daisy captain.

Who it is for — If you are an AgentTeams agent using DeepSeek Harness with GINKA desktop assistant and need to report task status to Daisy captain via voice, this plugin is suitable for you. If you do not participate in AgentTeams team tasks or prefer other reporting methods, you can skip it.

Watch out — From source install, need to manually allow the build script first. Sandbox test passed after installing into a fresh profile and registering with harness. Kokoro models of about 175MB need to be downloaded.

The verdict — I would install it if my work involves AgentTeams agents reporting task completion to the captain, because it supports Kokoro-82M-zh offline TTS.

What it is — dsh-action-outbox provides an output-commit boundary for DeepSeek Harness tool calls.

Who it is for — Users of DeepSeek Harness coding agents performing code development who need to review tool side effects should install it. Users who require distributed transaction commits across tools should not install it because it is not a distributed transaction.

Watch out — Sandbox test passed in a new profile and the plugin was registered into the profile by harness. No static check issues were found.

The verdict — I would install it because it implements the withhold until commit option without pretending unrelated external systems share an atomic transaction.