BigFish: Equipping DSH with a Fishing Knowledge Base
# Introduction The philosophy of DSH is "everything is a plugin," where capabilities are extended through skills. However, most tools in the skill ecosystem are developer-oriented. If you one day ask an agent, "The atmospheric pressure is low today; what bait should I use for crucian carp?", general models often provide generic answers because they lack structured fishing knowledge. The Agent Skills Open Standard provides a simple solution: package domain knowledge into an SKILL.md file and several data files, which are automatically loaded when the framework starts, eliminating the need to manually paste materials into prompts. BigFish, introduced below, is a fishing skill implemented according to this standard. # What is BigFish BigFish (kobenfang/bigfish) is a fishing assistant skill for DeepSeek Harness (DSH). The author...
Read MoreBigSeed: Let dsh capture fleeting thoughts and piece together your story
BigSeed is a DSH community skill maintained by kobenfang (npm package @kobenfang/dsh-bigseed, v1.0.0,
Read Moredsh-todo-freshness-guard: Keeping the todo list fresh in DSH long-running tasks
# Introduction Developers running long tasks with DeepSeek Harness (DSH) have likely encountered this state: the Agent is still running, tool calls are coming one after another, but the todo list on the panel is stuck at a point several minutes ago—completed items still show as incomplete, new blockers are not recorded, and it is impossible to determine the next steps. The longer the task, the more unreliable this panel becomes. Such issues were typically addressed by simply instructing the model in the prompt to "update in time," lacking mechanism-level constraints. **dsh-todo-freshness-guard** takes a different approach: it counts and alerts at the Harness layer and, when necessary, directly blocks ordinary tool calls until the model resubmits a complete todo list. Below is an introduction to this plugin. # What is dsh-todo-freshness-guard...
Read Moredsh-zhihu-dashboard: Add a Zhihu dashboard and seven agent tools to DeepSeek Harness.
# Introduction For agent development, Zhihu is often an indispensable information source: choosing topics requires checking the hot list, following answers to specific questions, and monitoring updates from certain authors. The existing approach involves switching to the browser, reading the content, and then pasting the link and summary back into the agent conversation, causing a break in information flow between two copy-paste actions. DSH's philosophy is "Everything is a plugin." The `dsh-zhihu-dashboard` introduced below brings this capability into DeepSeek Harness (DSH): it opens a Zhihu panel in the UI sidebar to handle viewing and tracking, while simultaneously registering seven native tools within the conversation to allow the agent to directly check the hot list and search for content. # What is it `dsh-zhihu-dashboard` is a DSH plugin maintained by kexuejin under the MIT license.
Read Moredsh-zpdf: Adding PDF Parsing, Translation, and Conversion Capabilities to DeepSeek Harness
# Introduction PDFs are an unavoidable part of agent development: papers, reports, and contracts often lose information regarding formulas, tables, and multi-column layouts when fed directly to models. Translating a PDF requires piecing together a pipeline for parsing, layout adjustment, and re-writing. DeepSeek Harness (DSH) adopts the philosophy of "everything is a plugin"; such high-frequency capabilities need not be built from scratch. Below, I introduce dsh-zpdf, a community plugin that encapsulates PDF parsing, layout translation, format conversion, and points management into DSH tools. # What is it? Maintained by komoai2026, the npm package is named `@kolmopdf/dsh-zpdf` and is licensed under the MIT License. One-sentence positioning: ZPDF tools for DeepSeek.
Read Moredsh-clipboard-history: A recallable clipboard history for DSH agents
# Introduction The system clipboard retains only the most recent item. If you copy a log in the terminal and then copy an error message, the previous content is lost. For AI agent developers, this means that when DSH attempts to handle "the content copied just now," it can only access the latest entry. `dsh-clipboard-history` solves this problem: it monitors the system clipboard in the background, records every text change to a persistent file, and exposes this data to the agent as a tool, allowing the agent to retrieve the history by ID at any time. The following section introduces the plugin's features, installation method, and configuration options. # What is this `dsh-clipboard-history` is a DeepSeek Harness (DSH) plugin authored by kuangre.
Read Moredsh-workspace-labels: Color and label DSH sidebar workspaces and sessions
# Introduction After running many projects with DeepSeek Harness (DSH), the workspaces and session list on the left accumulate increasingly. Visually distinguishing workspaces with similar names is inefficient. The common workaround is to change naming prefixes, but renaming affects path resolution and comes with a significant cost. Below is a lighter approach: `l-s-c/dsh-workspace-labels`, an unofficial DSH community plugin that adds color and text tags to workspaces and sessions directly in the original three-dot menu in the sidebar, without renaming or changing session content. # What is it? One-sentence positioning: A client-side plugin for organizing the left sidebar workspaces and sessions using colors and text tags. - Author: `l-s-c` - Version: 0.6.3 - License: MIT - Type: Unofficial
Read MoreFruit Pi: Connecting DSH Agents to Global Fruit Wholesale Prices
# Introduction When building agents, you often encounter questions like "How much is durian today?" or "What is the current market situation for cherries?" Models lack real-time data, so they either refuse to answer or provide outdated prices based on their training. To address this, you need to connect an agent to a data source that can retrieve current pricing. Fruit Pi is a DSH skill designed to do exactly that: automatically collecting daily wholesale prices for popular fruits and standardizing the conversion to RMB/kg. Once installed, the agent can directly answer questions about fruit prices. Below is an introduction to its features, installation, and usage. # What is Fruit Pi? Fruit Pi is a global fruit price collection and query tool that automatically gathers daily wholesale prices for popular fruits such as durian, lychee, cherries, and mangosteen. It supports quotes from both domestic and international markets and performs automatic conversion (RMB/kg). The repository is located at kobenfang/Frui
Read Moredsh-deepseek-balance: Real-time display of DeepSeek API balance at the bottom of the dsh Web sidebar
# Introduction When developing with DeepSeek Harness (dsh), the DeepSeek API is usually indispensable. To check the remaining account balance, one typically has to leave the current environment and check the platform's backend. `dsh-deepseek-balance` brings this capability into the dsh web interface: the balance is displayed directly at the bottom of the sidebar; hovering reveals details, and clicking forces a refresh. Below is an introduction to the usage and implementation of this plugin. # What is it? `dsh-deepseek-balance` is a DeepSeek Harness (dsh) plugin maintained by kyya, currently at version 0.1.0 and licensed under the MIT License. It displays the balance in real-time at the bottom of the web sidebar (above the settings button).
Read MoreEyes: Global Hotspot Monitoring and Market Impact Analysis Plugin for DSH
# Introduction When building DSH agents, a common issue arises: while agents can search for information, they often lack continuous awareness of global hot events—knowing "what happened when" and its implications for the industry and market often requires building the entire pipeline for data collection, grading, and push notifications from scratch. Eyes encapsulates this pipeline into a plugin: monitoring, event grading, impact analysis, and scheduled notifications are all ready to use; simply install and go. Below, I will introduce its features, installation methods, and typical usage. # What is Eyes? Eyes (kobenfang/eyes) is a global hot event monitoring and impact analysis tool maintained by kobenfang. It covers global situations, geopolitical conflicts, major policies, and innovative technologies that may impact the economy and market, and automatically analyzes transmission impacts across industries, exchange rates, and commodity supply chains. License
Read Moredsh-nav-pointer: Add a message pointer guide rail to the left side of the DSH Web Chat interface
# Introduction When using the DSH Web client for long conversations, there is a common minor annoyance: wanting to return to a previously sent instruction requires scrolling up screen by screen with the scroll wheel, making it increasingly difficult to locate messages as the number of messages grows. The `dsh-nav-pointer` plugin solves this navigation problem. It adds a message pointer rail to the left side of the chat area, where each user message corresponds to a short horizontal marker. Clicking the marker smoothly scrolls to the corresponding position, and hovering allows you to preview the message content. The following sections introduce the plugin's purpose, features, installation, and configuration. # What is it? The `dsh-nav-pointer` is a DSH (DeepSeek Harness) Web client plugin maintained by kongdexu, licensed under MIT, and currently at version 0.4.0. It works via `shell`
Read Moredsh-mac-vision: Enable DeepSeek Harness to perform OCR and visual detection locally on macOS
dsh-mac-vision is a macOS native vision plugin maintained by Kevoyuan (MIT open source), providing native OCR and visual detection capabilities for text-only models within DeepSeek Harness (DSH). By leveraging macOS's built-in Apple Vision framework, it can read local images and the clipboard.
Read Morebigfocus: DSH's Information Tracking and Change Reporting Plugin
bigfocus is a DSH skill (Agent Skill) developed by kobenfang that enables agents to continuously track product prices, celebrity updates, industry information, or custom metrics and automatically report them upon any change. It follows the Agent Skills Open Standard (SKILL.md + YAML frontmatter).
Read Moredsh-aurora-bg: Apply a DeepSeek official website-style dynamic aurora background to DSH Web UI
# Introduction When using DSH for agent development, `dsh web` is one of the most frequently opened interfaces. However, the default background style is quite plain, and staring at it for a long time can become tedious. Adding visual customization to the Web UI usually requires modifying the frontend code yourself, which involves a high barrier to entry and maintenance costs. `dsh-aurora-bg` solves this problem: it turns the dynamic background of the DeepSeek official homepage—featuring a deep blue fluid background, aurora light spots, and a dotted whale—into a DSH plugin. Once installed and the interface is restarted, the background takes effect automatically without the need to modify DSH's core code. # What is this? `dsh-aurora-bg` is a DeepSeek Harness (DSH) Web UI background plugin developed by KUN
Read MoreBigFood: Enabling the dsh Agent to Understand Fridge Ingredients
BigFood (@kobenfang/dsh-bigfood) is an "AI Refrigerator Butler" skill maintained by kobenfang. Adhering to the Agent Skills open standard (SKILL.md + YAML frontmatter) and licensed under MIT, it is usable across frameworks such as dsh, OpenClaw, and Claude Code. It encapsulates the workflow of "uploading ingredient images -> identifying ingredients -> recommending recipes" as a native capability of the agent, eliminating the repetitive need to write prompts manually each time. Core features include multimodal ingredient recognition, recipe recommendation, multi-image mashup analysis, fridge inventory management, and purchase reminders. To install, clone the repository to the user-level `~/.agents` directory (usable by all projects) or the project's `.agents` directory.
Read Moredsh-session-repacker: Repacks DSH session logs into the official two-frame format, opening old sessions about 6 times faster.
# Preface If you have used DSH for long sessions, you have likely encountered this issue: opening an old session takes several seconds to respond. The problem lies in the storage layer. DSH's JSONL session backend compresses each batch write into a standalone zstd frame and appends it to the file's tail. A long session can easily contain tens of thousands of frames; reading requires decompressing frame by frame. Furthermore, the backend does not support seek-based suffix reading, meaning the entire file must be decoded whether you need just the tail or the whole thing. The number of frames is determined by historical writes and cannot be shrunk afterward. `kkishapppy/dsh-session-repacker` addresses this issue by repacking scattered thousands of log frames into the official native format. Measured results show the maximum time to open a session reduced from approximately 661ms to approximately 126ms.
Read Moredsh-btw: Use /btw to bypass questions in dsh web without interrupting the main task
# Introduction When running long tasks with agents, you likely encounter this scenario: halfway through a task, you suddenly want to ask, "What was the name of that config file?" If you send the question directly to the main input box, it enters the main conversation, causing the currently running task to pause and process this interruption—a small question that could be answered in a few seconds disrupts the entire main thread. Claude Code solved this problem with the `/btw` command: questions are routed via a side channel, while the main thread continues running. `dsh-btw` brings this capability to the web interface of DeepSeek Harness (dsh). The philosophy behind dsh is "everything is a plugin," and `/btw` is a community-driven supplement to this. Below is an introduction to its positioning, features, installation, and usage. # What is dsh-btw?
Read Morebigplan: Turning product research into a DSH skill and outputting three tiers of product specification plans
# Introduction In DSH (DeepSeek Harness), adding capabilities to agents typically involves writing scripts or integrating toolchains. However, tasks such as product research are different: their core lies in analysis and structured output, where scripts offer limited assistance; what truly matters is a set of well-defined prompts. bigplan is an attempt to encapsulate this type of work as a skill. Below, we introduce its positioning, installation method, and usage. # What is it? bigplan is a skill maintained by kobenfang within the Agent Skills ecosystem. In one sentence, it is defined as: analyzing the market, technology, and supply chain based on product direction to output three sets of product specification plans (high, medium, and low-end). It follows the Agent Skills open standard (SKILL.md + YAML front...
Read Moredsh-whale-celebration: Trigger the DeepSeek whale to spray water at the bottom of the page for 3 seconds upon task completion.
dsh-whale-celebration is a DSH Web plugin (MIT licensed) designed to solve the problem of repeatedly refreshing to confirm task completion when long tasks run in the background. When the root agent of the user's direct conversation changes from "running" to "idle", a DeepSeek official whale spouting water jumps up from the bottom of the page and sinks after 3 seconds. The animation does not trigger for child agent completions, and the animation layer does not intercept clicks. The plugin consists of a Host and a Client: the Host runs inside the DSH process to listen for "agent/status" events and provide polling interfaces, while the Client runs on the browser side to poll every 0.9 seconds and play the animation. Installation requires only one command followed by a restart; the package declares `dsh.bundle`, which is automatically merged into the profile, requiring no manual configuration editing. Multi-tab compatible.
Read MoreBigLead: B2B Lead Mining in DSH
BigLead (kobenfang/biglead) is a DSH skill dedicated to B2B lead generation. It packages the standardized workflow of "searching for target companies—multi-channel cross-verification—extracting contact information—managing lead databases" into a plugin, suitable for B2B sales, market research, and competitive analysis. It adheres to the Agent Skills open standard and is compatible with frameworks such as OpenClaw and Claude Code. It requires Python 3 and internet search capabilities to operate.
Read Moredsh-lcx-codex: Completing Native Compact, Hosted Search, and Prompt Cache for DSH's GPT Responses path
When developing agents using the GPT/OpenAI Responses route in DeepSeek Harness (DSH), if the compression, Replay, and search stages do not align with the provider's native capabilities, issues such as mid-way implementation changes and repeated cache invalidations are easily encountered. DSH's philosophy is "everything is a plugin," and these gaps are perfectly suited to be filled by plugins. The `dsh-lcx-codex` introduced below performs exactly this: it takes over the DSH GPT Responses path to provide Native V2 Compact/Replay, Hosted Search, Prompt Cache, and continuous web operations.
Read Moredsh-trading-toolkit: A read-only market data toolkit for DeepSeek Harness agents (A-shares/US stocks)
# Introduction To integrate market data into an agent, the common practice is to register with a market data service provider, apply for an API Key, configure the Key into the environment, and then write custom logic to identify A-share and US stock codes. This process must be repeated for every new agent. If you also want the agent to participate in trading decisions, you must additionally address credential security and the risk of accidental orders. dsh-trading-toolkit consolidates this into a DSH plugin: it handles market data queries, ADX status signals, and simple backtest previews, all in a read-only mode. The following section introduces its features, installation methods, and typical usage. # What is it? dsh-trading-toolkit is a DeepSeek Harness (DSH) plugin maintained by kentleenot.
Read Moredsh-wallpaper: Set a custom wallpaper for DeepSeek Harness Desktop
## Introduction The default interface background of the DeepSeek Harness (DSH) desktop application is uniform, which can become monotonous after prolonged use. For DSH, which follows the "everything is a plugin" philosophy, such visual customizations do not require modifying the main program; a single plugin installation is sufficient to resolve the issue. The plugin introduced here is exactly that: **dsh-wallpaper**, by author keke050, licensed under the MIT License. Its function is to set a custom background wallpaper for the DeepSeek Harness desktop application. It supports three source types: presets, image URLs, and local uploads, and provides a transparency slider to adjust the wallpaper's display intensity. ## What is it? dsh-wallpaper is a DSH plugin. The current version is 0.1.0, and the client platform is Web.
Read MoreBig8: Add an AI Metaphysics Assistant to DSH (Feng Shui · Physiognomy · Bazi · Divination)
## Introduction The philosophy of DSH (DeepSeek Harness) is "everything is a plugin"; the framework itself is only responsible for scheduling, and specific capabilities are filled by skill packages. Most skills in the community are concentrated in engineering scenarios like coding and information retrieval. If you want the agent to smoothly handle some life-oriented requests—such as a user uploading a floor plan to ask about feng shui, or providing a birthday to calculate the Eight Characters—it is far more convenient to install a ready-made skill according to the Agent Skills standard than to write calendar conversion and prompts from scratch. Below is an introduction to Big8, maintained by kobenfang. ## What is this? Big8 is a skill package that adheres to the Agent Skills open standard (SKILL.md + YAML frontmatter), maintained by kobenfan.
Read Moredsh-simple-cli: Add a Claude Code-like terminal chat to DeepSeek Harness
# Foreword Developers who work in the command line for a long time often have this requirement: input a sentence directly in the terminal to see the model's response in a streaming format, the tool execution process in real-time, and manually confirm sensitive operations—instead of switching to a browser or assembling outputs from headless calls. DeepSeek Harness (hereinafter referred to as dsh) supports specifying a profile at startup (e.g., `dsh --profile cli`). The dsh-simple-cli introduced below is a CLI profile that runs entirely within the terminal: once installed, use the `dsh-cli` command to converse with the Agent, resembling the interface of Claude Code. # What is dsh-simple-cli
Read More