dsh-motion-memory: Local Memory Management Plugin for DeepSeek Harness
dsh-motion-memory is a memory management plugin designed for DeepSeek Harness (DSH), aimed at solving the context loss problem in agent workflows. It leverages conversation tracking, periodic summarization, and a four-layer memory structure to automatically consolidate conversation content into local memory documents, while supporting reference traceability and read-only protection. The plugin fully supports local storage and local models, with an emphasis on privacy and cost control. Installation requires a specific command and strict pairing with DSH version 0.2.0-rc.2.
Read Morehw-cola/dsh-dynamic-agents: DSH Web Dynamic Context Plugin
DSH Web provides an agent context rule management plugin that adds a docked control above the chat box. The plugin supports two levels of rule libraries, global and workspace, allowing users to select, add, or delete rules and inject them into the system prompt in real time. Once a session starts, the control is locked and must be unlocked to make changes. Installation requires the `dsh` and `pnpm` environments, and it has been tested to work with DSH `0.1.7-rc.1`.
Read MoreAgent DAG Workflow: A Host-Neutral Runtime for Agent DAG Workflows
`@gm-hz/agent-dag-workflow` is a host-neutral DAG workflow runtime designed to transform discrete tool calls in environments such as DeepSeek Harness or Codex into structured, persistable, validated, and replayable `WorkflowTemplate` JSON. It does not include model or credential centers; instead, it focuses on orchestration capabilities, supporting CLI operations, visual Canvas design, MCP gateways, Journal persistence, and multi-host environments. Developers can integrate it via an SDK or CLI to audit and restore complex agent workflows.
Read Moresatan9394/dsh-issue-triage: DSH Skill: Issue/PR Triage and Brief Generation
dsh-issue-triage is a workflow plugin for DeepSeek Harness designed to automate triage routing, status transitions, and briefing generation for issues and pull requests. It supports classification, state machine management, and report generation, making it suitable for automated community feedback workflows. Since official installation commands and specific examples are not provided, it is recommended to obtain it from the GitHub repository and carefully review the source code to ensure it meets your expectations.
Read MoreDeepSeek Harness Plugin: dsh-agent-project-sync
The DeepSeek Harness plugin `dsh-agent-project-sync` is intended to consolidate project directories from agent tools such as Codex and Claude Code into DSH for unified management. It automatically discovers and registers these directories as native Workspaces, and provides CLI and DSH tools for scanning, inventory management, and secure synchronization. It is constrained by safety boundaries: it reads only file-header metadata, runs in dry-run mode by default, and never deletes existing data. After installation, users can manage the project inventory through the command line or built-in tools, while paying attention to the required permissions and the fact that the inventory contains local absolute paths.
Read MoreDSH Session Collaboration
`dsh-session-collaboration` is a plugin for DeepSeek Harness (DSH) designed to enable bidirectional context passing across multiple sessions. It implements collaboration through the `session_delegate` model tool, Host Runtime, and Web Card. Core features include relay context injection into target sessions, card navigation, and a wait-for-reply mechanism. Configuration supports cross-workspace collaboration and timeout settings. When using it, note its inherent limitations, concurrency limits, and error-handling mechanisms.
Read Moredsh-web-search-openai: Integrate OpenAI Responses API Web Search for DSH
dsh-web-search-openai is a DSH plugin developed by flg1217, designed to integrate a search provider based on the OpenAI Responses API into `ctx.web` without modifying the DSH source code. The plugin achieves server-side retrieval by invoking the `POST /responses` interface and utilizing the native `web_search` tool, unifying structured results or message-level annotations into citable sources. It supports a Web settings card that allows users to configure the endpoint, model, API Key, and retrieval context; notably, the API Key supports input without echoing or reading from the `OPENAI_API_KEY` environment variable. Installation requires ensuring the DSH version is not lower than 0.1.0-rc.6 and proceeding via `dsh pl`...
Read Moredsh-econ-tools: DSH plugin for econometrics research
`dsh-econ-tools` is a DSH plugin maintained by Chaos-Hyper, licensed under the MIT License, and positioned as a comprehensive econometric research assistant. It aims to integrate scattered research steps into a continuous workflow, covering method guidance, data cleaning, variable selection, empirical analysis, robustness checks, and result reporting. Core functions include recommending models such as OLS, IV, and DID based on research questions; providing missing value handling and Python code snippets; supporting theoretical, data-driven, and machine learning variable selection; automatically generating multi-language code templates; performing seven-dimensional robustness analysis including parallel trends and placebo tests; and outputting statistical reports in both Chinese and English in Markdown or LaTeX formats. This plugin is suitable for developers and researchers using the DSH web environment via `dsh plu`.
Read MoreDSH Tea Break (dsh-whale-cards): Local Game Hall Plugin for DSH Web
`dsh-whale-cards` is a local game plugin for DeepSeek Harness (DSH) Web, designed to assist developers in low-distraction casual interactions while waiting for long tasks such as compilation or testing. Based on the MIT License, this plugin offers three original mini-games, including "Whale Cards Tea Break," and a companion character named "Lan Yin." It supports saving and resuming progress and can be collapsed at any time without affecting background tasks. The plugin provides two modes: Classic Mode, which uses local rules and deterministic AI without consuming tokens, making it suitable for offline, low-cost gameplay; and Agent Mode, which creates an independent session for each round. In this mode, "Lan Yin" calls the selected model to make valid decisions, consuming tokens and potentially involving session data reading. Additionally, the plugin supports non-blocking notifications to ensure that DSH task completions or confirmations do not disrupt the game focus. Data is primarily stored in...
Read Moredsh-task-notify: DSH Round End and Background Task Notification Plugin
`dsh-task-notify` is a lightweight plugin designed specifically for DSH Web to address the issue where users struggle to obtain timely status updates for tasks after leaving the page during long agent turns or background tasks. The plugin ensures users do not miss completed, failed, or terminated tasks by providing sound and browser tab title notifications upon task completion. Core features include playing a synthesized dual-tone alert (with different pitches for success and failure) when an agent turn ends or a background task enters the `completed`, `failed`, or `killed` states. If the page is in an inactive tab, the tab title is modified to provide a reminder, which is automatically restored upon switching back. The plugin features an anti-false-positive mechanism that only applies to turns actually completed in the current session, preventing interference caused by historical data refreshes. Installation is simple and supported via the npm registry.
Read Moredsh-rule-engine: Delegating AGENTS.md rules to DSH for execution
dsh-rule-engine is a DSH Rule Execution Engine v3 plugin that uses ~/.dsh/AGENTS.md as the single source of truth, transforming document rules into runtime objects that are checkable, executable, and auditable. It enforces rules through tool guarding, text detection, timing checks, and audit logs, supporting dynamic parsing so that adding, deleting, or modifying rules does not require rewriting the plugin. Its main features include the /guard command for managing status, rules, logs, and whitelists, support for automatic mounting of bundle specifications, and capabilities such as task contracts, anti-over-engineering mode, and LLM adjudication. When there is no AGENTS.md in the environment, the engine loads without errors and does not produce false positives. The installation method is to use the dsh plugin --profile web add dsh-rule-engine command or manually...
Read Morellm-adaptive: Add an adaptive (automatic routing) model entry point for DSH
llm-adaptive is an MIT-licensed DeepSeek Harness (DSH) plugin designed to solve the challenge of frequently manually switching models within a session due to varying request complexity. The plugin provides an "adaptive (automatic routing)" option in the DSH model selector, enabling intelligent backend routing through per-request complexity classification. Core features include: classifying each request into one of four levels—`low`, `medium`, `high`, or `critical`—using the `deepseek-v4-flash` classifier combined with session context and recent dialogue; supporting Sticky level protection to prevent unnecessary degradation during task execution; and configuring the routing chain entirely via `pool.json`.
Read MoreIntegrate Deterministic Engineering Workflow into DeepSeek Harness with mstar-harness
### 正式翻译（符合技术文档规范）： mstar-harness (Morning Star) is maintained by btspoony, and uses the TypeScript engine to enforce access control for functions such as status/lease/dispatch. The mstar-* skills are responsible for multi-agent delivery judgment. This article verifies the differences between the directory installation command and the dsh plugin --profile web add @mstar-harness/dsh recommended in the README, as well as the usage of `/iteration-start` and `/codebase-audit`, based on the directory page, GitHub README and v2.3.0 release notes. The plugin runs with the permissions of the current dsh process, and you should check the source code and MIT license before installation.
Read More