Every DeepSeek Harness plugin

What it is — dsh-custom-brand is a plugin for customizing the brand area in DeepSeek Harness Web UI. Replace whale logo and DeepSeek wordmark with local images, and edit HARNESS badge text. Changes persist in localStorage.

Who it is for — If you are using DeepSeek Harness Web UI and want to customize the brand appearance, this plugin is suitable. Users not running the DeepSeek Harness Web UI do not need it.

Watch out — Sandbox test passed in a new profile where it was registered by harness. No obvious issues found.

The verdict — Install it if you frequently change UI branding elements, since modifications are stored locally and survive restarts; otherwise it is unnecessary as it only offers browser-based persistence.

What it is — dsh-negative-ledger is a plugin that persists disproven paths from failed commands and missing files with their outcome and precondition evidence, automatically warning or blocking repeats until evidence changes.

Who it is for — Coding agents using DeepSeek Harness that frequently invoke bash or pwsh tools and need to prevent repeated failures on identical commands or paths should install this plugin.

Agents that do not execute shell commands or read files do not need this plugin.

Watch out — Sandbox testing reveals that installation may be blocked by pnpm due to ignored build scripts in better-sqlite3; from source installation requires manually permitting the build script to complete registration. The plugin activates through the bundle layer in DSH profiles.

The verdict — I would install it for DeepSeek Harness coding agents handling frequent tool call failures because it provides evidence-bound negative memory that auto-revokes blocks when preconditions change, with the requirement of manually allowing the build script during npm install.

What it is — Integrates uisfx sound effects into DeepSeek Harness for semantic task and button cues.

Who it is for — For users executing AI agent tasks in DeepSeek Harness who need sound effects for task status and button interactions. For users who do not perform any agent tasks, this plugin is unnecessary.

Watch out — Sandbox test passed, the plugin was installed in a new profile and registered to harness. No static check issues found.

The verdict — I would install it because it exposes the ctx.uisfx service API for integration by other plugins, but only when using dsh version 0.1.0-rc.6.

What it is — DeepSeek Harness Web plugin that stashes current pure-text input in per-session LIFO stack and restores via collapsible panel.

Who it is for — In DeepSeek Harness web sessions requiring temporary stash of current input to handle sub-tasks. For users who include images, attachments or file references in their inputs, since the plugin only supports pure text.

Watch out — Tested successfully in a fresh profile with successful harness registration. No obvious issues found.

The verdict — I would install it because it provides explicit save-restore for temporary inputs beyond DSH's built-in draft saving.

What it is — This plugin synchronizes skills, instructions, session history, and MCP servers between OpenAI Codex and DeepSeek Harness.

Who it is for — You should install it if you use OpenAI Codex and need to import its skills or session history into DeepSeek Harness.
Those who only run within DeepSeek Harness and do not involve Codex synchronization do not need it.

Watch out — Installation requires using insert lines in cordis.patch.yml instead of overriding to avoid duplicate loader entry id.
It executes shell commands. Sandbox testing succeeded in a fresh profile.

The verdict — I would install it because it registers Codex skills as first-class in DSH and provides real-time MCP mirroring, but only if you can handle the cordis.patch.yml configuration in the profile.

What it is — Provides IM gateway adapters for Feishu, WeCom, and Telegram to connect DSH agents.

Who it is for — If you want to engage in multi-turn conversations with your DSH agent via Feishu messaging apps, this plugin is suitable. Users who prefer to manage DSH profiles solely through the web GUI do not need to install it, since it focuses on messaging platform integration.

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

The verdict — I would install it because it enables per-chat sessions with a visual configuration card in the web GUI.

What it is — dsh-rss provides RSS/Atom subscription management and parsing tools for the agent.

Who it is for — If you are using a DeepSeek Harness agent that needs to subscribe to RSS/Atom sources and fetch parsed content to enrich model outputs, this plugin provides tools like rss_add and rss_fetch. If you need to manage local files rather than network RSS/Atom sources, this plugin is not suitable as it specializes in external subscription scraping and parsing.

Watch out — Sandbox test passed: installed in a fresh profile and registered by Harness. It will execute shell commands. No obvious pitfalls found.

The verdict — I would install it because it persists the subscription list to settings for agent tool calls across restarts.

What it is — dsh-hyperframes integrates five official HyperFrames by HeyGen skills into DeepSeek Harness.

Who it is for — When developing video-related features in DeepSeek Harness agents, this plugin supplies the HyperFrames skills. If you do not intend to copy skills to other agents supporting SKILL.md format, you can skip this plugin.

Watch out — Requires Node.js ≥ 22 and FFmpeg. Sandbox testing in a fresh profile confirmed successful registration to harness. It executes shell commands.

The verdict — I would not install it because the benefit is limited to users already in the DeepSeek Harness environment needing these specific skills.

What it is — This plugin enables DeepSeek Harness agents to send Markdown or plain text notifications to DingTalk groups via webhook with HMAC signing and zero runtime dependencies.

Who it is for — If you are using DeepSeek Harness agents to automate tasks and need to notify a DingTalk group about workflow status, this plugin is suitable. If your tasks are confined to local profile environments without external group chat integration, you can skip this plugin.

Watch out — Sandbox testing confirmed successful installation and registration in a fresh profile with no runtime issues detected. Configuration requires specifying the webhook in cordis.patch.yml, with optional secret. Only Markdown and plain text message types are supported.

The verdict — I would install it for its zero-dependency webhook integration to DingTalk groups for agent notifications, but only if you can securely configure the webhook and manage the secret.

What it is — It provides CalDAV calendar tools for DeepSeek Harness agents to list, create, update, delete, and search events with RRULE expansion for Google/iCloud/Nextcloud/custom endpoints.

Who it is for — When DeepSeek Harness agent needs to schedule events through calendar integration, this plugin can be used. If the agent does not require interaction with external calendar services, it can be skipped.

Watch out — Configuration requires providing an API key or token. Sandbox testing passed in a fresh profile. No obvious issues found.

The verdict — I would install it because it inserts a calendar configuration line in the profile to enable agent scheduling.

What it is — Provide Codex-backed web_search, image_gen, and image_vision tools for DeepSeek Harness.

Who it is for — Users needing web search and image generation in DeepSeek Harness with Codex login. Not suitable for those without ChatGPT/Codex login, since it reuses the existing OAuth state.

Watch out — Sandbox test passed: installed in new profile and registered with Harness. No obvious issues found. Requires Node.js >= 22 and valid ChatGPT login. Uses built-in https transport with no npm dependencies.

The verdict — I would install it for its specific tool implementations, but only if I already have a Codex login.

What it is — dsh-desktop-windowos provides a Windows desktop shell for DeepSeek Harness with tray and desktop shortcuts.

Who it is for — Users performing conversation tasks with DSH on Windows should install it because it exposes desktop_launch tool. Users who do not need tray integration need not install it because it depends on the tray for window restoration.

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

The verdict — If I needed the desktop_launch tool for DSH on Windows, I would install it because it auto-updates.

What it is — Adds microphone voice input to the DeepSeek Harness composer, supporting live transcription with browser Web Speech API.

Who it is for — Suitable for users needing microphone voice input in the DeepSeek Harness Web UI composer. If not using a browser that supports Web Speech API or not running in browser environment, users do not need to install this plugin.

Watch out — Sandbox test passed after installation in a fresh profile and was registered into the profile by harness. No obvious issues found.

The verdict — I would install it because it deduplicates repeated text and auto-continues long sentences with zero server components via browser Web Speech API.

What it is — This plugin provides cross-session project-scoped SQLite memory storage for DeepSeek Harness, with six-layer structure management and tool interfaces.

Who it is for — When using DeepSeek models for long-term tasks involving multiple projects, users who need to maintain project facts, user preferences, and cross-window memory will find it suitable. If handling only single short-term projects with limited memory needs, installation is unnecessary.

Watch out — Registration in profile cordis.patch.yml is required followed by restart. Sandbox tests confirm successful registration in fresh profiles. No runtime issues detected.

The verdict — I would install it because per-window nightly memory consolidation supports long-term multi-project workflows.

What it is — dsh-output-styles plugin enables runtime switching of model output styles via the /style command, with Claude Code outputStyles parity, per-session persistence in the output_style domain, and system prompt injection.

Who it is for — When conducting software engineering tasks with Claude models, users who need to control output format at runtime to ensure consistent replies are suitable for installing this plugin. If your work mainly relies on default output styles rather than custom Markdown libraries, users can skip it because it focuses on system prompt injection.

Watch out — No obvious issues found. Sandbox test passed in a fresh profile where the plugin is registered. Static check only indicates potential shell command execution but no risks observed.

The verdict — I would install it because it allows precise runtime control over model output style with session persistence, which is valuable for consistent coding sessions.

What it is — It is a local multi-agent message relay plugin for DeepSeek Harness, integrating HMAC-authenticated broker and plugin tools.

Who it is for — When running multiple independent Agents on the same host for collaboration, this plugin is suitable. If you focus on single-Agent reasoning chains without inter-Agent communication needs, you don't need to install this plugin.

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

The verdict — I would install it because it fills the gap in standardized inter-Agent communication for existing frameworks.

What it is — Provide live preview for editing user and built-in system prompts in a dedicated tab.

Who it is for — When you need to precisely adjust system prompt positions and override native components in the chat page, users who use the Prompt Studio tab. Users who do not care about automatically capturing non-dialogue message contexts can skip it.

Watch out — Sandbox test passed: the tab registers normally in a fresh profile. No obvious issues found in installation or runtime.

The verdict — I would install it because it automatically captures non-dialogue message contexts.

What it is — Through auto-reconnecting SSH tunnel running dsh web on remote Linux server, allocating and registering dedicated remote ports with audit view.

Who it is for — If you need to run dsh web on a remote Linux server for development work and access via local tunnel with dedicated port allocation and audit, this plugin fits you. If your work is limited to local machine without remote server resources, this plugin provides no local environment support.

Watch out — Sandbox test passed in a new profile and successfully registered. Requires providing API key or token. No obvious issues found.

The verdict — I would install it because it solves the multi-user remote dsh web port allocation and audit problem, provided there is SSH key login to the remote server.

What it is — dsh-model-failover provides two-level model circuit breaker failover for DeepSeek Harness.

Who it is for — When you need to handle model failures in the agent/request loop of DeepSeek Harness, this plugin is suitable for installation. If your task only involves session-title or compaction calls, you can skip it.

Watch out — Sandbox testing passes: the plugin installs and registers in a fresh profile. No obvious issues found.

The verdict — I would install it because it can record failures escaped from llm-retry into the circuit, but only if fallbacks are configured and state is process-local.

What it is — DSH cross-session task relay plugin providing task push, claim, complete and handoff summary.

Who it is for — Suitable when using DSH for multi-session collaboration with sub-agents needing cross-session task relay. Users of single-session DSH without task handoff need not install it.

Watch out — Requires manually allowing the build script when installing from source. Tested successfully in new profile and registered into DSH profile. No obvious pitfalls found.

The verdict — I would install it because it provides reliable cross-session task handoff, the premise being multi-session workflows with sub-agents.

What it is — This is an evidence-first operating kit with six read-only tools and five packaged skills, providing capability catalog, workflow plans, skill reader, memory search, repository audit, and release checklist.

Who it is for — It suits agents doing research, multi-agent work, benchmarking, or releases that need to back claims with evidence before stating memory, plans, benchmarks, or safe releases. It does not suit users who want quick integration of pre-packaged tools, as this package requires source installation and local roots configuration.

Watch out — Installation from GitHub source requires manually permitting the build script. Sandbox testing passed: it installs into a fresh profile and registers with harness. No obvious pitfalls found.

The verdict — I would install it because it enforces evidence before claims, but only if my workflow involves planning or auditing where I can review the evidence myself.

What it is — dsh-polyglot provides model switching capability for DeepSeek Harness to OpenAI-compatible endpoints.

Who it is for — When conducting sessions in DeepSeek Harness with the polyglot model and wanting to handle free tier rate limits, users are suitable for this plugin. Users who prefer using only the deepseek-official provider can set it up directly in their profile without needing this plugin.

Watch out — Sandbox testing succeeded with registration in a new profile. The static check indicates execution of shell commands, with no other issues identified.

The verdict — I would install it because it automatically falls back to the next provider in the chain when a free tier rate-limits.

What it is — DeepSeek Harness plugin providing multi-character Galgame conversation view.

Who it is for — When performing creative writing or task planning in the DeepSeek Harness Web profile, it suits users wanting role-based assistance informed by workspace tasks. Not suitable for users who do not want the plugin to scan recent workspace sessions.

Watch out — No obvious issues found, but installation from source requires manually allowing the build script.

The verdict — I would install it because it integrates Galgame interactions with Harness task events, but I need to manually allow the build script.