Every DeepSeek Harness plugin

What it is — DeepSeek Harness plugin: surveys local code and GitHub/npm for reusable implementations before development.

Who it is for — When developing new components with DeepSeek Harness model after clarifying requirements, survey for reusable implementations. If only handling minor modifications below 50 lines, the model workflow rules will directly skip the survey.

Watch out — Sandbox test passed: installed in a fresh profile and registered by harness. From source install, need to manually allow the build script. No obvious pits found.

The verdict — I would not install it because it requires using the enterprise template to disable remoteSearch in company policies.

What it is — This plugin adds turn navigation to the DeepSeek Harness Web UI for long conversations. It follows scroll to preview user turns on hover and jumps to them on click.

Who it is for — When working with long conversations in the DeepSeek Harness web interface that require quick navigation to specific turns, this plugin helps. Users who do not use the DeepSeek Harness Web UI do not need it because this plugin only registers to the conversation.chat.navigator slot.

Watch out — Installation requires manually allowing the build script from source. It was tested to register successfully in a new profile. No obvious issues found.

The verdict — I would install it for users handling extended conversations in DSH web to save navigation time, but only if they must use the DSH Web UI and need the navigator feature.

What it is — dsh-trace exports DeepSeek Harness session telemetry to a local embedded yiTrace database, recording turns, model steps, and tool calls.

Who it is for — When using a DeepSeek Harness web or headless profile to handle agent turns and require detailed logging of model reasoning and tool calls, this plugin can be installed. Users who prefer HTTP-based telemetry services or do not need local trace storage may find it unnecessary.

Watch out — Installation requires sourcing from GitHub with manual build approval and a writable data directory. Sandbox testing confirmed successful registration in a new profile. No obvious issues detected.

The verdict — If you need full local observability of agent traces on your machine without an HTTP telemetry backend, I would install it because it supports direct queries via the yiTrace API.

What it is — Provides per-agent tool discovery and progressive schema disclosure.

Who it is for — 适合在 agent 需要动态补充工具列表来处理任务时使用。 不适合在已有完整工具列表无需搜索的项目中安装。

Watch out — Sandbox test passed after fresh profile install and registration. No obvious issues found.

The verdict — I would install it because it provides per-agent tool discovery, but only if accepting the unreleased experimental plugin with no compatibility promise.

What it is — Durable token budget management for DSH agent-tree, reserving capacity before llm/stream attempts.

Who it is for — When running concurrent child agents in a DSH profile, needing to manage token budgets to prevent all from spending the same remaining balance. If workflow is only single session linear execution, the durable descendant tree scope is unnecessary.

Watch out — Sandbox test passed after installation in a new profile with successful registration. The profile must provide Storage, Storage Domain, and Token Meter services in manual composition. No obvious pitfalls discovered.

The verdict — I would install it because it offers durable scopes and absolute deadlines for agent-tree token management, but only if the profile supplies the required storage services.

What it is — It is an A-share financial data MCP server for DeepSeek Harness, integrating 11 fin_data__* tools for unified access over THS primary + AKShare fallback + Wind exclusive domains, with an offline 10-year market data lake.

Who it is for — Chinese A-share retail investors are suitable for installing it because they need AI-assisted investment decisions and unified financial data access with DSH.

Users who do not use DeepSeek Harness need not install it because they do not require MCP integration for financial services.

Watch out — Installing from GitHub source requires Python 3.12+, uv sync, configuring THS and Wind keys in env or ~/.dsh/.credentials.yaml, then patching with cordis.patch.yml to integrate the MCP server. Sandbox test passed, but the plugin executes shell commands. Measured context budget compression reduced result-side by 72.2%.

The verdict — As an A-share retail investor using DSH, I would install it because it provides an offline data lake and observable degradation, but only if I supply my own API keys and accept shell execution.

What it is — Provides secure Git tools for models including status, diff, log, commit, and branch.

Who it is for — When the model needs to manage a code repository in a development session. Not suitable for sessions requiring stronger isolation.

Watch out — Sandbox test passed: installed in a new profile and registered by Harness, running normally. Git tools run as host process without fs-sandbox isolation. Commit requires git user.name and user.email configured in the repo.

The verdict — I would install it because it makes git calls safe and passed sandbox test, but with the premise that the session allows host-level git execution.

What it is — This is a DeepSeek Harness plugin for hatching and evolving a pet.

Who it is for — If you use a DeepSeek Harness profile and want a pet that evolves based on your work, install this plugin. It is not suitable for users needing multiple independent pets across different profiles, due to the single-pet-per-home limit.

Watch out — From-source installation requires manually allowing the build script. Sandbox test passed: installed into a new profile and registered by harness. No obvious issues found.

The verdict — I would not install it, because a single DSH home can only support one pet.

What it is — dsh-model-router is a routing policy plugin for DeepSeek Harness that adds deterministic keyword routing, allowlisted model switching, and an isolated image generation channel.

Who it is for — In DeepSeek Harness for product design and frontend planning, users needing model switching and image isolation should install it. Users who do not require complex reasoning or model switching do not need to install this plugin.

Watch out — Installing from GitHub source requires executing build steps, which may need manual allowance of the build script. Sandbox testing passed and it registered successfully in a new profile. No obvious installation issues found.

The verdict — I would install it because it routes models by keyword without guessing and isolates the image generation channel.

What it is — This DeepSeek Harness plugin binds your ChatGPT account via official OAuth and enables chatting with ChatGPT models inside DSH using your subscription quota.

Who it is for — When you need to chat with ChatGPT models in the DSH model switcher for conversations, this plugin provides official OAuth binding. If you do not plan to add the ChatGPT provider to your DSH models, you do not need to install this plugin.

Watch out — Sandbox test passed on a fresh profile; the plugin was installed and registered by Harness. No obvious issues found.

The verdict — I would install it because it binds via official OAuth and keeps existing login intact.

What it is — DeepSeek Harness Web UI integrates a run game where you can play while the agent replies, with progress strip displaying scores from 梁子 to 梁圣.

Who it is for — When using DeepSeek Harness Web UI for agent conversations, users wanting to insert a run game during replies are suitable for installation. Other users who use non-web profiles of DeepSeek Harness can skip it, as the plugin targets the web profile configuration.

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

The verdict — I would install it because it allows playing a run game during agent replies, but only if already using the web version of DeepSeek Harness.

What it is — DeepSeek Harness tool plugin evolution controller supporting signals to solidify.

Who it is for — When you need to trial custom tool plugins in DeepSeek Harness to handle tool errors, this plugin is suitable to install. If you prefer using official provided tools rather than developing custom plugins, this plugin is not applicable.

Watch out — Sandbox test shows it can be installed but the build script is blocked by pnpm, leading to harness not completing registration. Users need to manually allow the build to install properly. No other obvious pitfalls found.

The verdict — I would install it because it provides the evolution loop for tool plugins from signals to solidify, but installation requires manually allowing the build.

What it is — pack-agent packs skills, rules, and MCP into .pack.zip and installs to multiple harnesses.

Who it is for — When developing projects with Claude Code and needing cross-tool skill sharing, it fits. It is not suitable for users who only operate locally without multi-harness needs.

Watch out — Requires Bun ≥ 1.1. Sandbox testing succeeds in a fresh profile with successful harness registration. No obvious issues found.

The verdict — I would install it because it unifies skill management across supported harnesses, but only if using those tools.

What it is — dsh-webgate enables remote access to the DSH web GUI, including LAN QR access, cloudflared quick tunnels, and frp reverse proxying via your own server, with a built-in auth portal.

Who it is for — Need to access DSH web GUI from external devices. If you do not want to run cloudflared or frp processes, you can skip it.

Watch out — Sandbox testing passed: it was installed in a fresh profile and registered by harness into the profile. Static checks indicate it will execute shell commands and require API key or token. No obvious issues found.

The verdict — I would install it because it supports frp reverse proxying via your own server for stable external access.

What it is — Adds pause, resume, and cancel functionality for conversation tasks in the composer area, with configurable pause granularity.

Who it is for — If you are using the composer area with a model for conversation tasks and need to control tool execution, this plugin provides pause options. If you are not using the DSH profile system, you do not need it.

Watch out — Sandbox test passed in a new profile and successfully registered by harness. It will execute shell commands. No other obvious issues found.

The verdict — I would install it because the pause state persists after restart.

What it is — Injects CSS to provide mobile adaptation for the DSH Web UI with compact typography and composer fixes.

Who it is for — When using the DSH Web UI on a touch device, safe touch targets and full-width composer are needed. Desktop users on wide screens do not need it because the plugin only activates on touch or narrow viewports.

Watch out — Sandbox test passed: successfully installed in a new profile and registered with DSH. No obvious issues found. Requires restarting DSH web and browser hard refresh after installation.

The verdict — I would install it because it offers pure CSS injection with zero dependencies that can be uninstalled anytime.

What it is — Provides Web Components support as an installable profile bundle exposing a browser-side webComponent service that mounts registered slot trees as native Custom Elements.

Who it is for — If you need to mount slot trees as native Custom Elements in the browser environment of DeepSeek Harness to display model responses, this plugin is suitable for you. If your workflow does not involve browser-side slot integration, you do not need to install it.

Watch out — Installation from GitHub source requires manual approval of the build script to register. The bundle adds a disabled web-component row that must be explicitly enabled in the profile after the slot host and browser runtime are present. No other issues were observed.

The verdict — I would install it because it enables browser-side Custom Element mounting from slots, but I would only proceed if the profile already includes the required slot host and runtime and I am willing to manually allow the build script.

What it is — It extracts the Microsoft MXC sandbox provider from the DSH feat-mxc implementation as a standalone plugin for cross-platform support.

Who it is for — If you need to run DSH bash and PTY plugins in DeepSeek Harness for cross-platform shell command execution involving Windows or macOS environments, this plugin supplies the required sandbox capability. If your work does not require cross-platform compatibility or operates in stable single-platform environments, you may not need this plugin.

Watch out — The build script is blocked by pnpm during GitHub source installation, so from source install you must manually permit the build script or harness registration will fail. It executes shell commands, but no other issues were found.

The verdict — I would install it for cross-platform sandbox needs in my DSH workflows because it integrates Microsoft MXC support, but only after manually permitting the blocked build script for proper registration.

What it is — DSH Web settings bundle integrating host configuration crawler and browser navigation.

Who it is for — When editing DSH settings through the browser page in your profile, install this plugin for web-config-crawler support. Users relying on gateway RPC configuration can skip it.

Watch out — Sandbox test shows it installs but pnpm blocks the build script from github:omdsh-dev/ex-setting, preventing harness registration. Manual approval of the build is required for full registration. No other issues found.

The verdict — I would install it because it targets browser Web configuration only, but only after manually allowing the pnpm build script.

What it is — It converts documents to auditable SQLite FTS5 knowledge packs supporting deterministic retrieval and original text精读.

Who it is for — It is suitable in the DeepSeek Harness agent knowledge pack construction scenario because it provides pure computational kb_build, kb_query, kb_read tools with zero sub-agents, zero workflow, zero external services. It is not necessary to install if using the original Python kb-sieve because the knowledge packs generated by this plugin can be queried directly.

Watch out — Sandbox test passed, installed in fresh profile and registered into harness profile. Static check found it will execute shell commands. No other obvious pits found.

The verdict — I would install it because it generates pure data skills without scripts or Python dependencies and is compatible with original knowledge packs.

What it is — DeepSeek Harness desktop pet plugin featuring official whale outline with agent state animations.

Who it is for — When using DeepSeek Harness in a web profile for agent tasks, this pet provides state animation accompaniment. Users who prefer only plain text interfaces without any desktop elements won't need it.

Watch out — Sandbox test passed, successfully installed and registered in a fresh profile. No obvious issues found. It requires DSH version >= 0.1.0-rc.6 to operate.

The verdict — I would install it because it has zero runtime dependencies and passed the sandbox test.

What it is — folio-tools is a DSH plugin providing 15 schema-validated tools for memory and quality gates plus an L0 guard.

Who it is for — When using DeepSeek Harness for consulting material generation and needing memory management plus quality gates, folio-tools is suitable. Users who only rely on CLI tools without DSH plugin integration can skip it.

Watch out — Requires PowerShell and Python ≥ 3.10. Sandbox testing not performed. No obvious issues found.

The verdict — I would install it because it supplies DSH-native schema-validated tools and L0 guard for long-term task execution.

What it is — The folio-events plugin provides session-protocol events for DSH, with auto session-start reminders and auto-save on session close.

Who it is for — DSH users running long-term tasks who want session events can install it. Users not running DSH sessions don't need it.

Watch out — No sandbox test performed yet. No issues found in static checks.

The verdict — I would install it if I need the session events to enhance DSH plugin layer management, as it's not required for basic use.

What it is — The plugin allows DeepSeek Harness assistants to automatically send emoji packs when they have emotions, with library management and AI image tagging.

Who it is for — Users who chat with an AI assistant in a DeepSeek Harness web profile and want it to proactively send emoji packs based on emotional states. Users who prefer completely passive conversations without the assistant proactively analyzing and sending images can skip this plugin.

Watch out — The sandbox test passed with successful registration in a new profile. Plugin data is stored locally and does not upload chat records. AI image tagging calls a configured model service.

The verdict — I would not install it because installing from source requires manually enabling the build script first.