dsh-plugin-judge: DSH Plugin Value Judge
`dsh-plugin-judge` is a plugin value judgment tool in the DeepSeek Harness ecosystem, designed to address the issue that a plugin's value can fluctuate with model changes. It provides three core capabilities: pre-installation review (static scanning + LLM judgment), post-installation audit (risk scoring), and reminders for re-evaluation when switching models. Through commands or the settings panel, users can assess plugin suitability and manage constraint-related risks, ensuring that plugins do not impair task performance due to rule restrictions or model upgrades. Installation requires environment prerequisites such as Node.js >= 18; judgment results are for reference only, so install third-party code cautiously.
Read MoreDeepSeek Harness Subagent Model Routing Plugin: Automatically Route Flash and Pro
The dsh-delegate-router plugin addresses the excessive costs caused by DSH subagents inheriting the parent model by default (such as Pro). It automatically routes lightweight tasks to the Flash model and keeps demanding tasks on the Pro model based on task characteristics, with support for custom rules, manual overrides, and decision logging. The plugin takes effect without a restart, significantly reducing inference costs through model routing.
Read Moredsh-balance: DeepSeek Harness Sidebar Balance Monitoring
dsh-balance is a DeepSeek Harness plugin designed to display account balances accurately in real time through the official API endpoint. It uses SSE long-lived connections to push data and features low-balance alerts, bilingual support in English and Chinese, and zero dependencies. Users need to configure environment variables and alert thresholds in the configuration file. The plugin offers a simple interface that displays real data in the account's currency, making it suitable for developers who need precise monitoring of API resources.
Read Moreskk-gal: A Sikke Theater View Plugin for DeepSeek Harness
`skk-gal` is a Skirk-themed web chat view plugin for DeepSeek Harness (DSH). It uses a modular design and provides an immersive experience with no network dependency. Core features include a Skirk theater view (with character-art breathing effects, multi-turn card display, and task panel integration), a fullscreen adaptive theme, draggable sidebar resizing, and rich chat display modes. Installation is deployed locally via a PowerShell script and supports mobile adaptation. Users can manage typing speed, character animations, and chat style in settings. Archived sessions support online deletion and recovery from the local recycle bin.
Read Moreheadless-: CI Structured Output Plugin for DeepSeek Harness
`headless-` is a plugin designed for DeepSeek Harness (dsh) to enhance the reporting capabilities of CI pipelines. By subscribing to the session event stream, it transforms interactions into structured data and provides features such as JSON/NDJSON transaction reports, JUnit XML test reports, semantic exit codes, and file artifact collection. The plugin supports privacy masking, ensures deterministic output without interfering with standard output, and is suitable for automated evaluation scenarios in agent development.
Read MoreRepoGate: GitHub Developer Workbench and MCP Toolkit
`repogate` is an open-source MCP tool plugin designed to wrap the GitHub REST API into standard interfaces, enabling agents to directly manage repositories, issues, PRs, and code reviews through conversation. It offers zero runtime dependencies, supports both PAT and OAuth authentication, and includes features such as a built-in read-only mode and concise output. After installation, models can invoke related tools using clients such as `dsh` to achieve efficient interaction. The plugin requires Node.js >= 18.17 and provides detailed error messages in Chinese.
Read Moredsh-mcskin: A Minecraft Theme for DeepSeek Harness
dsh-mcskin is a third-party, Minecraft-style plugin for DeepSeek Harness. By overriding theme tokens and injecting CSS, it provides six theme configurations featuring sidebar gradients, brand grass strips, Creeper decorations, and pixelated icon controls. Theme selection supports light/dark mode and is persistently saved via the browser's LocalStorage. The plugin requires a recent DSH version and modern browser support, is maintained by community developers, and aims to deliver an immersive, playful interface experience.
Read MoreDeepSeek Harness Plugin: Integrating OpenAI Codex (ChatGPT Plus/Pro OAuth)
`ddll8023/dsh-codex` is a DeepSeek Harness (DSH) plugin designed to connect to OpenAI Codex models via OAuth without requiring a DeepSeek Key. It supports browser and headless login, with features such as native web search, Fast mode switching, and automatic credential refresh. Users can manage login, usage, and session modes through CLI commands, securely storing credentials in DSH's own service.
Read Moredsh-access-mode: DeepSeek Harness Session Access Mode Plugin
`dsh-access-mode` is a plugin for DeepSeek Harness (DSH) designed to address the limitations of native permission preset balance and the high cost of modifying core code. By providing three session-level access modes—Default / No Edit / Auto—it enables users to finely control the AI’s tool invocations, sandbox policies, and approval policies without modifying the core code of Harness. **Core features**: non-intrusive integration, reuse of native services, and session persistence support. **Mode descriptions**: Default uses a conservative `workspace-write` boundary; No Edit removes file restrictions while only restricting editing; Auto fully permits access. **Security note**: No Edit and
Read Moredsh-cross-session-agent: DeepSeek Harness Cross-Session Agent Coordination Plugin
`dsh-cross-session-agent` is an unofficial plugin for DeepSeek Harness designed to address conflicts and coordination issues in parallel multi-agent development. Through peer discovery, message passing, transport receipts, and restricted context projection, it enables agents to securely negotiate file ownership, report progress, and request handoffs. Its core capabilities include discovering peers, sending coordination directives, querying transport status, and reading restricted contexts. Strict permission boundaries must be maintained during use: only regular Sessions in the same workspace as the current one may be read, and reading sensitive reasoning data or executing commands contained in messages is prohibited. The plugin provides the infrastructure for multi-agent collaboration through its safety mechanisms.
Read Moredsh-notify Plugin: Desktop and Remote Browser Notifications
dsh-notify is a plugin for DeepSeek Harness (DSH) designed to address the issue of delayed monitoring of agent task status. It provides desktop toast notifications and remote browser push alerts, supporting real-time reminders when a task is completed, encounters an error, or awaits user intervention. The plugin supports command-based control, a global toggle, and session filtering. It relies on Windows and WinRT to implement desktop notifications and is compatible with DSH version 0.1.7.
Read Moredsh-theme-dodger-17: DeepSeek Harness Dodger Blue Stadium Theme
The DeepSeek Harness plugin `dsh-theme-dodger-17` provides a customized Dodger Blue-based visual scheme for the DSH Web UI. User-maintained, it includes hand-tuned light/dark color schemes and supports one-click switching in DSH settings, automatically remembering the user's preference. Technically, it runs purely on the web client by overriding documented tokens and is compatible with both the DSH Web UI and Desktop versions. Users can install it by downloading the `.tgz` file and using the `dsh plugin add` command, then enabling it in settings.
Read Moredsh-codebuddy-auth: Using Tencent CodeBuddy as a Native LLM Provider for DeepSeek Harness
dsh-codebuddy-auth is a plugin that integrates Tencent CodeBuddy (IOA) into DeepSeek Harness, providing a native adapter and an out-of-the-box experience. It supports browser-based OAuth login, automatic token renewal, and model list synchronization, and exposes inference capabilities based on `/v3/config`. The plugin uses a dual-identity mechanism to bypass authentication restrictions and supports access to the international version. After installation, users can directly switch to CodeBuddy models in the DSH model selector, while the system automatically handles credentials, streaming, and model discovery.
Read Moredsh-proxy: Injecting a Custom HTTP Proxy into DeepSeek Harness
dsh-proxy is a plugin developed for DeepSeek Harness (DSH) to address its default behavior of not reading system proxy settings or environment variables. By injecting a ProxyAgent to intercept global fetch traffic, it supports mainstream proxy tools such as Clash and mihomo, enabling all network requests to be routed through a proxy without enabling TUN mode. The plugin uses the Clash mixed port (127.0.0.1:7890) by default and offers flexible configuration, including host, port, and a noProxy bypass list, making it particularly suitable for environments where IP addresses are blocked or network access is restricted.
Read Moredsh-metrics-panel: DeepSeek Harness Usage Monitoring Panel
`dsh-metrics-panel` is a monitoring dashboard plugin designed for DeepSeek Harness, addressing common developer pain points such as tracking token consumption, costs, and latency. The plugin provides features including real-time statistics, visualized trend charts, multi-dimensional aggregation, and billing configuration, with support for dark mode toggling and historical data refresh. After installing it via the CLI, users can view usage and costs across different time dimensions in the Web interface sidebar. Note that the data is stored in memory and cleared after a restart; the `dsh.bundle` declaration must be checked during installation.
Read MoreDeepSeek Harness Plugin: dsh-vector-memory
`dsh-vector-memory` is a persistent memory plugin in the DeepSeek Harness (DSH) ecosystem. It aims to address agents' cross-session memory needs with zero external dependencies by leveraging DSH's built-in `storageDomain`. The plugin provides three core tools — `mem_save`, `mem_search`, and `mem_health` — and `vectorMemory` service injection, supporting real-time display of memories on the right side of the client interface. Installation only requires adding `@bryan-cmf/dsh-vector-memory` in the configuration file. Note that this plugin was merged into `dsh-insights` on 2026-08-17; please follow that repository for future iterations.
Read MoreDeepSeek Harness Plugin: dsh-skill-trail
dsh-skill-trail is a DeepSeek Harness (DSH) plugin designed to render compact skill and tool usage trajectories beneath each assistant reply, using engine logs rather than model self-reports. It supports structured data display, replay, and collapsing, and depends on DSH version `>= 0.1.0-rc.6`. It has since been superseded by `dsh-skill-ledger` and archived, with only design history records retained.
Read MoreDeepSeek Harness Plugin: dsh-skill-gate Skill Triggering and Selection Gate
`dsh-skill-gate` is a DeepSeek Harness plugin designed to address issues of model skill-selection bias and unclear intent. It intercepts the skill dispatch process and returns the final decision to the user. The plugin provides a native UI to display candidate skills, supports a 10-second timeout with automatic fallback to the default skill, and is compatible with headless mode. Note that this plugin has been superseded by `dsh-skill-ledger`, and only one gate can be enabled at a time.
Read MoreDSH-Plugin: Structural Observability Layer
The `dsh-infra-observability` plugin for DeepSeek Harness (DSH) is designed to provide a structured observability layer. It records actual tool execution, agent state, and watchdog alerts, instead of relying on model self-reports. Core features include `tools/result` listening, health monitoring, and a browser dashboard. The plugin offers three query tools: `usage_report`, `audit_skills`, and `infra_health`. Installation requires configuring the package name and a DSH version `>= 0.1.0-rc.6`. Its companion package is `dsh-skill-trail`; this suite has since been merged into the `dsh-insights` repository.
Read Moredsh-infra-bootstrap: Infrastructure Readiness Probing and Status Reporting
`dsh-infra-bootstrap` is a DSH plugin developed by Bryan-cmf that automatically probes the status of infrastructure components before task execution to avoid missed manual checks. It provides the `infra_status` and `infra_bootstrap` tools, which check components such as vectorMemory, search services, and skill directories to output a structured readiness report (PASS/WARN). The plugin requires DSH version >= 0.1.0-rc.6 and helps users quickly verify environment connectivity before starting an agent workflow.
Read Moredsh-recipe: A Recipe Aggregation Tool for Plugin Environments
dsh-recipe is a tool in the DSH ecosystem for packaging plugins into scenario recipes. It addresses users' need for complete development environments through natural-language search, merging, and one-click deployment. The plugin supports browsing, searching, expanding, and merging community-provided recipes, with 10 common scenarios built in. Its core advantage is that it runs entirely locally, providing the security benefits of zero dependencies, zero network requests, and zero writes. Users can install it with the `dsh plugin` command and manage environments using the `recipe` tool (such as list, apply, and compose).
Read Moredsh-plugin-verify: Evidence Verification Toolkit for DeepSeek Harness Agents
`dsh-plugin-verify` is a read-only workflow plugin for DeepSeek Harness (DSH), designed to help agents verify their behavior and environment state through automated checks. It provides multiple verification modes: keyword-based evidence checks, configuration file validation (with JSON/YAML structure testing), HTTP resource probing, NPM package and GitHub repository checks, and batch concurrent checks. The plugin relies on heuristic search, does not support write operations, and is primarily used in code auditing, configuration validation, and resource probing scenarios.
Read Moredsh-plugin-recommend: A DeepSeek Harness Plugin Discovery and Recommendation Tool
dsh-plugin-recommend is a local plugin search and recommendation engine embedded in DeepSeek Harness (DSH). It includes 1,135 built-in plugin entries, supports keyword, tag, and category-based search, and can recommend alternatives based on similarity, with real-time refresh and schema validation. Developers can quickly locate tools via five commands: search, list categories, view details, find similar, and refresh. Schema compatibility and security scanning should be verified before use.
Read Moredsh-repo-atlas: DeepSeek Harness Repository Analysis and Review Plugin
`dsh-repo-atlas` is a plugin for DeepSeek Harness designed to address the lack of structured evidence and change review boundaries for AI agents in unfamiliar codebases. It provides repository analysis, evidence-first conclusions, and bounded change proposals with lifecycle management. The plugin runs in read-only mode by default, strictly limiting the scope of operations to ensure safety, and supports installation through build and configuration. Developers should note that controlled actions are disabled by default and should review the source code before installing.
Read MoreDeepSeek Harness Prompt Optimization Plugin (zzy-dsh-prompt-optimizer)
The `zzy-dsh-prompt-optimizer` plugin is designed for DeepSeek Harness (DSH) to streamline the iterative optimization of prompts. It integrates into the DSH Web interface by adding a “Prompt Optimization” entry in the input box. The plugin supports three optimization modes (fidelity, developer-oriented expression, and specification expansion) and can access session and workspace context. Core features include directly populating the input with optimized results, one-click restoration of the original prompt, and configuration of model and character budget settings. It does not store API keys and only temporarily submits drafts for processing, with a focus on privacy and security. By using visualized interactions, the tool reduces the operational overhead for developers and is suitable for development scenarios that require frequent instruction tuning.
Read More