Every DeepSeek Harness plugin

What it is — dsh-doctor provides DeepSeek Harness diagnosis, repair with rollback, and runtime self-healing service.

Who it is for — When using agent in DeepSeek Harness web profile, if plugin causes crash, diagnosis and repair can be done after installing dsh-doctor. If only want to avoid handling build permissions, can skip it, because the build script is blocked by pnpm during install.

Watch out — Sandbox test shows it can be installed via npm but the build script is blocked by pnpm, preventing DeepSeek Harness from completing registration. Manual approval of the build is required to fully install. No other obvious issues found.

The verdict — I would install it because of the reversible repair and runtime self-healing features, but must manually allow the build script first, otherwise it won't complete registration.

What it is — Provides self-healing runtime layer plugin for DeepSeek Harness.

Who it is for — When cluster level trust collaboration and cross-session knowledge carry-over is required, this plugin is suitable. For users without need for error-storm detection, this plugin's adaptive capability is not relevant.

Watch out — Sandbox test passed, successful installation and registration in a fresh profile. It will execute shell commands. No obvious pitfalls found.

The verdict — I would install it because it provides the complete cybernetic closed loop.

What it is — dsh-apple-mode supplies an Xcode AI integration mode for DeepSeek Harness sessions. It adds 26 Xcode MCP tools, 10 Apple platform skills, and an Xcode Intelligence-style persona via an agent preset that activates only when Apple Mode is selected.

Who it is for — On macOS, users developing Apple platform applications with DeepSeek Harness who lack Xcode integration tools and Apple skills are suitable for installing it. Users not developing Apple applications with Xcode do not need it because they do not require these integrations.

Watch out — Installation from GitHub requires running install.sh and selecting an Xcode version. Sandbox testing passed: it installs successfully into a new profile and registers with harness. No obvious issues found.

The verdict — I would install it because it provides dedicated tool integration for Apple development, but only if macOS and Xcode environments are available; otherwise it is not worth it.

What it is — dsh-memoria integrates the memoria vector-graph memory backend into DeepSeek Harness, providing observe/remember/search/recall tools for agents.

Who it is for — If you are developing agents in DeepSeek Harness that need to remember user facts and recall them in subsequent conversations, this plugin is suitable. If you do not plan to configure badge tokens via environment variables, you can skip installing this plugin.

Watch out — Sandbox test passes: installed in a new profile and registered into the profile by harness. Requires running a separate memoria service and providing badge token via environment variable. No obvious issues found.

The verdict — If I were you, I would install it because it enables persistent memory for DeepSeek Harness agents, but the premise is running the local memoria server.

What it is — dsh-model-sync syncs live model lists from providers into DeepSeek Harness settings and displays 5h / 7d plan windows or metered balance for the current session model.

Who it is for — When using the DeepSeek Harness composer and tracking quotas for the session model, this plugin is suitable. Users who manually manage model lists in the settings page without composer can skip it.

Watch out — The sandbox test passed in a fresh profile. The plugin registered correctly and activated on install. No obvious issues were found.

The verdict — I would install it because it integrates quota info directly into the composer for the session model, but only if using DSH version with the composer class names from 0.1.0-rc.6.

What it is — dsh-devtools is a metadata-first runtime profiler plugin for DeepSeek Harness agents.

Who it is for — When you use DeepSeek Harness to analyze agent turn and step execution metadata, you can use it. If your sessions are mainly for quick summaries rather than needing execution metadata, you can skip it.

Watch out — From source, manually permit the build script first. Sandbox testing passed; it installs and registers in new profiles. Compatible with DeepSeek Harness v0.1.0-rc.6. No obvious issues found.

The verdict — I would install it for its detailed execution metadata viewing, but only if my workflow requires that analysis rather than quick summaries.

What it is — DeepSeek Harness Web's session log daily token usage dashboard.

Who it is for — Users tracking token consumption and cost estimates from session logs in DeepSeek Harness should install this plugin. Users not using the DeepSeek official provider can skip this plugin.

Watch out — Sandbox test passed after installation in a new profile and registration by harness. Zero model token usage. No obvious issues found.

The verdict — I would install it because it supports time-aware pricing schedules and does not pollute statistics for custom gateways.

What it is — dsh-sideband provides asynchronous context relay for DeepSeek Harness. It freezes a bounded session snapshot, summarizes it asynchronously with an isolated no-tool LLM, and delivers a provenance-bearing capsule to another session or room without interrupting the active agent.

Who it is for — When you are handling complex tasks in one DeepSeek Harness session and need to pass focused context to another session or room without waiting or interrupting, this plugin fits your use case. Users who do not need cross-session context transfer do not need to install it, as they can complete all interactions within a single session.

Watch out — Installation requires sourcing from GitHub and requires Node.js ^22.19.0 || >=24 together with DeepSeek Harness 0.1.0-rc.6. Sandbox testing passed in a fresh profile. No obvious issues found.

The verdict — I would install it because it supports safe background context handoff without disturbing the current agent turn, which is valuable in collaborative multi-session workflows.

What it is — Adds pre-approval to tool calling in DeepSeek Harness.

Who it is for — When you use DeepSeek Harness to perform tool calls, this plugin allows manual approval before each execution. If you do not use tool calling, this plugin is not applicable.

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

The verdict — I would install it because it can control which tools require approval, but the premise is that the user configures the include and exclude lists for tools.

What it is — DSH Testkit provides Docker-isolated real-host lifecycle testing for DeepSeek Harness plugins.

Who it is for — Developers testing a plugin's install boot register update uninstall process are suited to install it. Teams doing only static checks do not need it because the testing dimensions differ.

Watch out — Sandbox testing passed: installed in a fresh profile and registered by harness. Requires Node.js 22+ and Docker, will execute shell commands. No obvious pitfalls found.

The verdict — I would install it because it provides structured evidence, but it does not prove that arbitrary executable code is safe.

What it is — dsh-power-button provides a one-click restart or stop button pinned to the bottom right corner of the DeepSeek Harness window via detached helper.

Who it is for — When you need to quickly restart DeepSeek Harness without closing the session, this plugin is suitable.

If your workflow does not involve GUI button controls or command-line management only, you do not need it.

Watch out — It has been tested to register successfully in a new profile and pass sandbox checks.

From source install, manual permission for build script is required.

It executes shell commands.

The verdict — I would install it because detached helper enables restarts, but only after handling build permissions.

What it is — It delivers a one-click launcher for DeepSeek Harness Web UI on Windows, double-click to boot the server and open the browser, plus web_launcher model tool.

Who it is for — Windows users running DeepSeek Harness Web profile want quick access to Web UI via launcher. Users who do not integrate the web_launcher model tool do not need this plugin.

Watch out — Sandbox test showed installation failure due to package.json starting with UTF-8 BOM, which JSON.parse rejects. No other obvious issues found.

The verdict — Currently cannot install due to UTF-8 BOM issue in package.json, so do not try it directly; wait for the fix and reassess based on needs.

What it is — This plugin provides 12 tested engineering skills. Every SKILL.md passes a format validator aligned with the official skill rules plus an isolated load smoke test.

Who it is for — Users who need git commit, code review, test-first, plugin dev, debugging, changelog, PR review, task breakdown and other engineering skills when using DeepSeek Harness can install awesome-dsh-skills. Those who only want to use existing engineering tools without expanding the skill list are not suitable for this.

Watch out — It requires installation from GitHub source. Static check indicates it will execute shell commands. Sandbox test passed: installed in a fresh profile and registered into harness.

The verdict — I would not install it because it only provides a verified list of skills rather than directly integrating engineering tools.

What it is — In the DeepSeek Harness Web GUI settings, it provides toolchain inventory, cache cleanup, and AGENTS.md editor.

Who it is for — When running DeepSeek Harness agent sessions that require toolchain location and version checks, the plugin is suitable for installation. Users who do not use the DSH web GUI settings panel do not need this plugin.

Watch out — Sandbox test passed: installed in a fresh profile and registered with harness. No runtime dependencies were found.

The verdict — I would install it because it supports automatic backup of the previous version when editing AGENTS.md.

What it is — dsh-acp-enhanced is an enhanced Agent Client Protocol server for DeepSeek Harness.

Who it is for — Users of DeepSeek Harness in Zed who need real-time streaming output and telemetry can use it. Users who only want the official plain text streaming ACP bridge do not need it.

Watch out — The sandbox test passed with successful registration in a fresh profile. No obvious pitfalls found.

The verdict — I would install it because it overcomes the official bridge's plain text streaming limitation.

What it is — This plugin ports the Matt Pocock skill set to DeepSeek Harness Cordis plugin architecture, registering a skill provider at the host layer of ctx.skills and providing 25 skills including grilling, writing-for-agents.

Who it is for — When you need to merge grilling, writing-for-agents or code-review skills into the scope chain of an agent preset in DeepSeek Harness, install this plugin. Users who do not need to register skill providers into the profile scope chain can skip it.

Watch out — Sandbox test passes after installation registers into profile. Static check flags shell command execution but this is not a defect. No obvious issues found.

The verdict — I would install it because it mounts skills to the profile scope chain via import.meta.url, but only if you need those skills.

What it is — Register the read-only tool repo_setup_scan, which scans the project directory for language stack, test settings, documentation, git, Docker, and database clues, then provides recommended plugins, MCP servers, and hygiene files.

Who it is for — When the model enters a new or unfamiliar repository, users are suitable to install it to automatically invoke repo_setup_scan. Users who do not need to provide recommended plugins are not suitable to install it, because they already have complete configurations.

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

The verdict — Install it because it provides an automatic scan report for recommendations, but only if configuration suggestions are needed.

What it is — Delivers local-first structured vision for text-only DeepSeek Harness agents: images route to a local OpenAI-compatible VLM and return JSON evidence with summary, verbatim OCR, layout regions, entities/relations, colors, and explicit uncertainty.

Who it is for — It fits users handling image-inclusive tasks with text-only models that lack native vision support. Agents with built-in image capabilities do not require this plugin.

Watch out — Successfully tested in a fresh profile with harness registration. Depends on a running local VLM endpoint. No obvious issues found.

The verdict — I would install it for local-privacy vision on text agents with existing VLM access, as it enforces structured evidence and anti-hallucination fallback; otherwise, not worth it.

What it is — Renders Mermaid code blocks in conversation messages into interactive diagram cards with zoom, fit-width, fullscreen viewer, and preview/code toggle using official UI primitives.

Who it is for — When using DeepSeek Harness Web UI for conversations containing Mermaid code blocks, this plugin is suitable for installation. Users running DeepSeek Harness on non-web client platforms do not need this plugin.

Watch out — Sandbox test passed with installation in a fresh profile and successful registration. Static check indicates it will execute a shell command. No other issues detected.

The verdict — I would install it for agent conversations in the DSH web UI that include Mermaid diagrams because it enables interactive rendering, but only after handling build script approval for source installs.

What it is — Creates a Windows desktop shortcut for DeepSeek Harness web to launch the server and open the browser when ready, and adds a draggable floating refresh button.

Who it is for — If you need to launch DeepSeek Harness web on the Windows platform, this plugin is suitable.

Watch out — Sandbox test passed, in a new profile it is installed and registered into profile. The plugin will execute shell commands to launch and stop services. No obvious issues found.

The verdict — I would install it because it automatically stops the service when the Harness window closes.

What it is — dsh-memory provides cross-session memory storage, supporting remember/recall/forget tools, per-turn system prompt injection of the latest entries, and a settings page browser to manage the records.

Who it is for — When using DeepSeek Harness for long-term projects requiring persistent context across sessions, this plugin allows agents to retain project information and inject it automatically. Users who avoid shell command execution entirely can choose other plugins without this feature.

Watch out — Sandbox testing confirmed successful installation in a fresh profile and registration with the harness, with no obvious pitfalls found. The plugin executes shell commands, which represents a real cost.

The verdict — Given the need for cross-session memory, I will install the plugin assuming shell command execution is acceptable in my environment.

What it is — DeepSeek Harness plugin detects garbled text features before AI writes and requires review before release, plus pre-execution pwsh encoding audit.

Who it is for — When using pwsh tools to write Chinese files on Windows, this plugin is suitable for installation. Users who do not execute shell commands can skip this plugin.

Watch out — Sandbox test passed after installation in a new profile and registration. It will execute shell commands.

The verdict — I would install it because it provides a specific tradeoff.

What it is — KeyringSeam provides a signed notarized macOS Keychain credential provider for DeepSeek Harness replacing the local-file provider using a universal helper.

Who it is for — Users who need credential management in DeepSeek Harness web profiles are suitable for installing KeyringSeam. Users running on macOS below 13 should not because it targets macOS 13 or newer.

Watch out — Sandbox testing passed in a fresh profile with successful registration. No obvious issues found. Building from source requires manually allowing the build script.

The verdict — I would install it because it delivers agent isolation via signed notarized helper, but only on macOS 13+ after accepting authentication prompts.