Every DeepSeek Harness plugin

What it is — peanut-dsh-plugin is a DeepSeek Harness plugin collection.

Who it is for — Users who want to install personal plugins with DeepSeek Harness are suitable for this collection. Users who do not need to execute shell commands need not install this collection.

Watch out — Sandbox test passed: installed in a new profile and registered to harness. It will execute shell commands when installing from GitHub source.

The verdict — I will install it because the sandbox test passed and it can be installed from source.

What it is — This is a tool that packages DeepSeek Harness as a cross-platform Electron desktop application.

Who it is for — Users who want to run DeepSeek Harness in a desktop window format can install it on macOS Windows Linux platforms. People who only run Harness in a browser and do not need desktop integration do not need it.

Watch out — Installation from source involves a build step. Sandbox testing passed after installing in a new profile and registering it with Harness. It will automatically execute scripts and run shell commands during install. No obvious issues found.

The verdict — I would install it because it keeps upstream sync and adds desktop integration but only if the user accepts the source build step.

What it is — This is a DeepSeek Harness plugin. It allows text-only models to receive images in conversations. Images are saved to the session workspace and handed to the model as file paths.

Who it is for — If you use a text-only model in a DeepSeek Harness profile for conversations requiring image input, this plugin is suitable. If you use a multimodal model supporting native image attachments or only perform text-based interactions, you do not need it.

Watch out — Manual approval is required for the build script when installing from source. Sandbox testing confirmed successful registration in a new profile. No obvious issues were found in operation.

The verdict — I would install it because it enables image handling for text models via file paths without affecting multimodal models.

What it is — Integrates claude-mem memory management into DeepSeek Harness for cross-session memory query, context injection, memory save, and session summarization.

Who it is for — Users of DeepSeek model for cross-session dialogue are suitable for installation because they need mem_search tool and mem_context. Those not needing to run claude-mem worker do not need to install it.

Watch out — GitHub source installation requires manually adding allowBuilds to pnpm-workspace.yaml to allow the prepare script. Sandbox test passed, registered into profile. No other issues found.

The verdict — I would install it because it integrates claude-mem via local HTTP API and passed sandbox test.

What it is — DeepSeek Harness plugin integrating SSH remote workspaces for file operations and command execution.

Who it is for — When using the DeepSeek Harness agent to develop or debug on remote servers that require file access and command execution, this plugin offers support. Users who only rely on local workflows can continue using the default local providers.

Watch out — Sandbox test passed: the plugin was installed in a fresh profile and registered by Harness. Static checks indicate it will execute shell commands, requiring attention to security risks. No other obvious issues found.

The verdict — This plugin is worth installing for its transparent integration of remote operations with local providers, but only if your workflow involves SSH to remote hosts.

What it is — Delivers Death Stranding skin and live heartbeat feed for DeepSeek Harness UI.

Who it is for — Suitable for users who use DeepSeek Harness agent for thinking and tool execution because it allows agent activities to have heartbeat wave feedback. Not suitable for users who do not use session snapshots because it relies on session snapshots to drive the heartbeat.

Watch out — Successfully installed and registered into profile in sandbox test. No obvious issues found.

The verdict — I would install it because it provides heartbeat feedback on agent activities but the premise is needing real-time data driven display.

What it is — dsh-vision-helper registers the vision_analyze tool to analyze images and provides a configuration interface.

Who it is for — Developers using DeepSeek Harness for image analysis tasks can install it. Users who only use text-based models can skip this plugin.

Watch out — Sandbox testing passed: it was installed in a fresh profile and registered with harness. No obvious issues found.

The verdict — I would install it because of its zero-dependency design and support for configurable multimodal models.

What it is — This is a DeepSeek Harness WebUI plugin that adds file management and change review interface to the right sidebar in the session page.

Who it is for — When using DeepSeek Harness Web Client in the current session workspace, this plugin is suitable for people who need file tree browsing and Git diff review. Users who do not use DSH Web Client can skip installing it.

Watch out — Sandbox testing shows it can be installed but the build script is blocked by pnpm, requiring manual approval of github:LoftyTao/dsh-ui-workbench. No obvious issues found.

The verdict — I would install it because it provides file management and Git review, but manual build script approval is required.

What it is — DeepSeek Harness Web UI plugin.

Who it is for — If you need a browser interface to interact with DeepSeek Harness agent, I would consider it. If your environment does not have a valid DeepSeek API key, I would not recommend it.

Watch out — Sandbox test shows it can be installed but requires manual approval of the build to complete registration. No obvious issues found.

The verdict — I would not install it because sandbox test shows it requires manual approval of the build to complete registration.

What it is — This plugin controls the DSH service residency mode for the dsh-launcher shell, allowing selection between follow-window, tray residency, or always-on.

Who it is for — Users running DSH services through DeepSeek Harness who want to save memory by stopping the service on window close should install this plugin. Users not running services through the dsh-launcher shell need not install it.

Watch out — Sandbox testing passed with the plugin registered in a fresh profile. No issues were found in static checks. No obvious pitfalls detected.

The verdict — I would not install it because the control only applies when using the paired dsh-launcher shell and has limited standalone value.

What it is — DASH is a terminal TUI plugin of DeepSeek Harness, providing agent interaction for the official dsh CLI.

Who it is for — When you are in a terminal environment using the dsh CLI with DeepSeek Harness for agent session management and require multi-line input editing, undo operations, and oh-my-pi style keybindings, DASH is suitable. If you do not need the kitty keyboard protocol and SGR mouse support provided by DASH, you do not need to install it.

Watch out — Installation requires GitHub source installation and running the install script first. Sandbox test passed in a new profile, successfully registered into the harness profile.

The verdict — I would install it because it provides complete terminal support for DeepSeek Harness's official agent services.

What it is — DSH plugin integrates tokensave semantic graph via CLI direct connection, providing semantic graph tools.

Who it is for — When agent needs to perform semantic queries to replace file operations, it is suitable to install this plugin. If the agent does not want the plugin to execute shell commands, it does not need to be installed.

Watch out — Sandbox test passed: installed in a new profile and registered to harness. Static check indicates execution of shell commands, but no other defects found.

The verdict — If I were a developer, I would install it because it automatically follows the working directory after installation without manual setup.

What it is — Provides a mobile-responsive AppFrame for the DeepSeek Harness web UI, adding drawer sidebar and bottom sheet for <640px width.

Who it is for — Suitable for users of DeepSeek Harness on devices under 640px width, as they need drawer sidebar and bottom sheet. Not suitable for users on devices over 640px width, since the Wide/Narrow behavior is reused from upstream.

Watch out — Sandbox test passed: it was successfully installed in a fresh profile and registered with harness. No obvious issues found. Old-kernel browsers (pre-ES2024 WebView) cannot load the client bundles.

The verdict — I would install it because it provides mobile UI adaptations for narrow screens, but only if using a current WebView.

What it is — A DeepSeek Harness plugin providing Claude Code-style terminal interaction.

Who it is for — Developers who want to do coding in the terminal are suitable to install this plugin. Users who prefer browser-based development can choose DeepSeek Harness's Web UI.

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

The verdict — I would install it because it provides a full-screen terminal UI, but only if you have Node.js 22 or later and an OpenAI-compatible provider.

What it is — dsh-paste-input enhances DSH WebUI file input, supporting paste, drag and drop, and file selection. Attachments are copied to the session workspace temp directory when sending, and conversation bubble text blocks are folded into file chips.

Who it is for — When using DSH snapshot0812 for session input, users who need paste input enhancement are suitable to install this plugin. Those users who do not use specific slots like conversation.input.dock are not suitable to install this plugin because this plugin targets specific slots.

Watch out — Sandbox test passed: installed in a new profile and registered into profile by harness. Compatible with multiple DSH snapshots and npm versions, no cordis dependency. No obvious pits found.

The verdict — I would not install it because it only passes sandbox test and registration on supported snapshots and npm versions.

What it is — The DSH bash output encoding auto-detection plugin replaces ctx.bash to collect raw bytes and auto-detects multiple encodings for decoding.

Who it is for — Windows users running bash commands under WSL NAT mode with proxy config who see Chinese garbled output. Users on Windows native profiles using pwsh do not need it.

Watch out — Compatible with DSH snapshot0808, snapshot0809, and the npm release. All test scenarios passed including mixed encoding streams. It registers correctly into a new profile. Default disabled in Windows native profiles.

The verdict — Will install it because it fixes garbled Chinese output from bash tools in WSL scenarios where real tests confirmed correct decoding.

What it is — DSH Web UI plugin, a pixel whale stays resident in the session header right side and reacts in real time to conversation snapshots with blinking, tail wagging, fin movements, thinking motion, water spray, heart bubbles and sleep.

Who it is for — Users who need dynamic decoration added to the session header in a DeepSeek Harness Web UI conversation are suitable for installing it. If you do not want any animations in the session header, you can skip it.

Watch out — When installing from GitHub source code you must select the build corresponding to the DSH snapshot version. When installing from source you must first manually whitelist its build script. It executes shell commands. Sandbox testing passed; it installs into a fresh profile and is registered by harness. No obvious issues found.

The verdict — If I need real-time whale decoration in the session header I would install it because it requires zero core changes and sandbox testing passed.

What it is — dsh-ui-progress plugin provides a persistent session progress bar in the DeepSeek Harness Web UI input dock area, showing real todos progress and real-time token generation rate.

Who it is for — Users who need to monitor session progress while using DeepSeek Harness Web UI for multi-turn dialogue generation are suitable for installing this plugin. Users who do not use DeepSeek Harness Web UI for dialogue are not suitable for installing it because they do not need progress visualization in this specific interface.

Watch out — Sandbox test passed: installed into a new profile and registered by harness, with progress states functioning normally. Builds require matching the corresponding DSH snapshot version for compatibility. Source installation requires manually allowing the build script.

The verdict — I would install it because it delivers real-time token rate without touching agent-loop, with sandbox test passing.

What it is — dsh-tool-hashline provides hash-anchored read/edit/grep tools for DeepSeek Harness.

Who it is for — When using DeepSeek Harness agent for code editing tasks with risk of post-read file changes, not suitable for users who do not want to create a preset.

Watch out — npm install succeeds as ordinary dependency without becoming profile layer. Shell commands will be executed. From source install requires manual approval of build script.

The verdict — I would install it to handle anchor verification during edits, but must create the preset as described in README.

What it is — DSH plugin list page adds bilingual descriptions to plugin cards and publishes pluginDescriptions service.

Who it is for — When you need to view plugin card functions in DSH Web settings plugin list page, this plugin fits. If you do not use DSH Web profile, you need not install it.

Watch out — Sandbox test passed: installed in new profile and registered by harness. Static check indicates it will execute shell commands.

The verdict — I will install it because it supports description registration by other plugins and custom persistent mounting, but only if you use the web interface for plugin list management.

What it is — dsh-delegate extends DeepSeek Harness subagent tools with per-call model selection, dependency gating and persistent run tracking.

Who it is for — When performing multi-step subagent delegations in DeepSeek Harness and needing to assign different models to steps or track dependencies, dsh-delegate is suitable to install. Users who only need one-time subagent calls without dependency management do not need to install it.

Watch out — Installation requires adding tool instances to the profile's cordis.patch.yml file. Sandbox testing passed with no obvious issues found, and it executes shell commands on runtime.

The verdict — I would install it because it offers deterministic model routing and dependency enforcement for complex delegation workflows.

What it is — Provide real-time attention alerts for DeepSeek Harness: a global banner plus flashing workspace entries when the agent is blocked by approval request or sandbox denial.

Who it is for — If you're running AI agent tasks in DeepSeek Harness and need to monitor sessions in real time for approval requests or user interactions, this plugin is suitable. Users who do not require constant visibility into agent blocking events may skip it.

Watch out — Manual build permission is required when installing from source because of the intercepted build script. Sandbox testing shows it integrates but needs user approval for registration. No other obvious issues found.

The verdict — I would install it because it offers practical alerts that shorten wait times for agent actions, but only after manually allowing the build script.

What it is — WhaleKit is a SKILL.md skills framework and five-tier routing system for DeepSeek Harness.

Who it is for — When doing agent work in DeepSeek Harness that requires step-by-step task processing and skill invocation management, this fits. People who do not involve any agent processes can choose other approaches.

Watch out — Sandbox test passed after installing into a new profile and it being registered by harness. No obvious issues found. Static checks note that it will execute shell commands.

The verdict — I would install it because the user remains the ultimate decision maker and the agent never decides the tier alone.

What it is — dsh-project-mcp-bridge plugin defines MCP servers per project via .dsh/mcp.json, automatically registering tools in project sessions with live reload.

Who it is for — It suits you if you need the model to call tools like GitHub API in a specific project. It enables per-project custom setup. Users relying only on preset MCP servers without needing project-specific differences do not need it.

Watch out — Sandbox testing passed: registers successfully in a fresh profile. Static analysis indicates execution of shell commands and requirement for API key or token.

The verdict — I would install it because per-project configuration with live reload reduces complexity when projects require different MCP servers.