Every DeepSeek Harness plugin

What it is — Sidebar usage and cost estimation panel for DeepSeek Harness.

Who it is for — Want to track context usage and cumulative cost when using DeepSeek Harness for multi-turn conversations. People who don't want to see achievement badges don't need to install it.

Watch out — No obvious pitfalls found. Data processing is entirely local with no network requests.

The verdict — I would install it because it provides official billing estimates for reference, while the final billing is based on DeepSeek platform statements.

What it is — DSH-changeproof plugin binds to DeepSeek Harness and automatically links tests to code changes, row-level coverage checks, and expires stale conclusions.

Who it is for — When using DeepSeek Harness to handle code changes and verify that modified lines are covered by tests, this plugin is suitable to install. If code changes are made without verification of test coverage or non-vitest/jest/pytest testing frameworks are used, no need to install.

Watch out — Requires Node.js ≥24 and pnpm for install. Sandbox test passed: installed and registered in fresh profile. Execution uses argv-only mode; no obvious issues found.

The verdict — I would install from source because it provides independent CLI and auto-calls changeproof_verify when bound to DeepSeek Harness.

What it is — Adds Net Access mode to DeepSeek Harness, solving the issue of curl being unable to access HTTPS in Windows sandbox while keeping file write protection unchanged.

Who it is for — When using DeepSeek Harness in a Windows sandbox environment to perform network operations requiring curl for HTTPS requests, this plugin can be installed. Users not using the Windows sandbox or primarily relying on system tools can use DeepSeek Harness on other operating systems.

Watch out — Sandbox testing passed, allowing normal HTTPS access while maintaining file write protection. However, the system curl cannot access HTTPS and must use the tools provided by the plugin. The plugin is bound to DSH 0.1.0-rc.6 and requires reinstallation after upgrading DSH.

The verdict — If I were you, I would install it because it solves the issue of curl being unable to access HTTPS in the Windows sandbox, but only if using DSH version 0.1.0-rc.6.

What it is — This is a frontend plugin for DeepSeek Harness providing theme, keyboard shortcuts and history features.

Who it is for — If you frequently switch models in complex conversations with DeepSeek models and need custom keyboard shortcuts to improve efficiency, this plugin suits you. If you mainly use a single model and don't care about history or usage panels, you can use the native interface directly.

Watch out — Sandbox testing passed: installed in a new profile and registered into Harness. No obvious issues found.

The verdict — I would install it because it provides practical customizations for model switching while keeping zero shell modifications.

What it is — This is an unofficial theme plugin that adds two virtual rooms for Joi to DeepSeek Harness.

Who it is for — Fans of virtual streamer Joi who want to see her character performances in the chat interface should install this plugin. Users who prefer a plain text interface without visual elements can skip it.

Watch out — Installation requires Node.js and pnpm. It passed sandbox testing. No obvious issues found.

The verdict — I would install it because it adds character expressions and automatic text highlighting to the chat interface.

What it is — DeepSeek Harness sticky notes plugin.

Who it is for — When using DeepSeek Harness web mode and DeepSeek model to read and write notes, this plugin applies. Users not using DeepSeek Harness web mode do not need it.

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

The verdict — I would install it because it is open source.

What it is — DeepSeek Harness plugin registers ocr_image tool to convert image paths to text output.

Who it is for — Use it when the agent needs to read text from image paths in TUI or Web profiles. Users who never have image paths in agent inputs do not need this plugin.

Watch out — Installation requires Python packages and model download script. Sandbox test passed after registration in new profile. No obvious pitfalls found.

The verdict — I would install it because it enables the agent to handle agent-accessible images offline.

What it is — Directly integrate Claude Code's .claude/ directory into DeepSeek Harness for zero-migration reuse of skills, commands, and rules.

Who it is for — When developing a project that follows Claude Code conventions in a specific DeepSeek Harness profile containing .claude/skills/, .claude/commands/, and .claude/rules/ directories, the plugin enables direct invocation of those skills and commands. If your workflow relies only on DeepSeek Harness's built-in skills and rules without needing to bridge external directories, you do not need this plugin.

Watch out — Sandbox test passed with successful installation and registration in a new profile. The static check noted execution of shell commands, which is a prompt rather than a defect. No other issues identified.

The verdict — I would install it because it provides a direct channel for integrating Claude Code configurations into DeepSeek Harness, but only if the project follows the .claude/ directory structure; otherwise, DeepSeek Harness's native tools suffice.

What it is — This is a DeepSeek Harness plugin that integrates the HeartFlow engine to provide 47-dimensional discrimination tools and automatic output supervision.

Who it is for — If you generate content with DeepSeek Harness and need to intercept hallucinations before users see them, this plugin fits. If your tasks do not require content verification tools, you do not need to install it.

Watch out — From source installation you must manually allow the build scripts, which sandbox testing shows pnpm blocks for protobufjs@6.11.6 and sharp@0.32.6, preventing full registration. No other obvious issues were found.

The verdict — I would install it because it provides automatic output supervision, but manual build script approval is required for registration.

What it is — Integrate Feishu bots into DeepSeek Harness via QR code scan.

Who it is for — When you need to converse with the Harness agent in Feishu, this plugin fits the scenario. If you prefer browser-based management of Harness, you can skip installing it.

Watch out — Sandbox test passed: installed in a new profile and registered successfully. The plugin runs in the dsh web Host process without needing extra services. No obvious issues found.

The verdict — I would install it because it allows independent management of multiple bots without cross-interference, but only if you need multi-bot support.

What it is — Integrates AtomGit skills, ag CLI, and platform-hosted MCP tools into DeepSeek Harness.

Who it is for — DeepSeek Harness users who need to manage AtomGit code repositories through issue planning, PR implementation, review, and merge are suitable for installation. Users who do not use any platform integrations or shell commands are not appropriate for this plugin.

Watch out — Sandbox test passed in a new profile with registration by harness. From source install, need to manually allow the build script. No obvious issues found.

The verdict — I would install it because all three modules share one unified AtomGit PAT with no separate token setup required.

What it is — Feishu bot chat bridge to DeepSeek Harness agent responses.

Who it is for — Feishu chat users who want to interact with DeepSeek Harness agents need to configure allowlist for responses. Users of other chat tools do not need this plugin.

Watch out — Sandbox test confirms successful installation as ordinary dependency. However, package.json lacks dsh.bundle so harness treats it as regular dependency and does not become a profile layer. This is not installation failure, it is that it is not originally a bundle-type plugin. No obvious pitfalls found.

The verdict — I would install it because sandbox confirms it installs successfully as ordinary dependency, but it does not become a profile layer and requires following the README to set environment variables.

What it is — Adds shared real browser plugin for DeepSeek Harness, enabling agent to drive visible native Electron window shared with humans for observation and takeover.

Who it is for — When performing web automation tasks with DeepSeek Harness agent that require human oversight, this plugin is suitable to install. If your tasks do not involve visible browser control, using headless solutions is sufficient.

Watch out — No obvious issues found. Sandbox test passed: installed in new profile and registered into profile. Self-hosts in plain dsh web environment without desktop shell.

The verdict — I will install it because it self-hosts on plain dsh web without requiring a desktop shell.

What it is — This is an OpenAI-compatible NewAPI LLM provider plugin that registers a newapi route in DeepSeek Harness, enables chat-only model discovery with auto parameter filling from models.dev, and adds a web settings section for base URL and API key configuration.

Who it is for — If you need to integrate a custom API gateway for LLM chatting in DeepSeek Harness, you can use this plugin to add the newapi route. If your current model selection already meets your needs, you do not need to install it.

Watch out — It passed sandbox testing in a fresh profile and was registered by harness. It executes shell commands. No obvious issues found.

The verdict — I choose not to install it because it only implements an OpenAI-compatible gateway; if you already have a suitable LLM provider, it may not be worth adding this plugin.

What it is — dsh-loop is a DeepSeek Harness plugin that provides recurring loops using a /loop command, a loop tool and an activity status bar, supporting multiple loops in parallel.

Who it is for — When you use DeepSeek Harness for polling, PR review, or build-fix-test cycles, dsh-loop is suitable for you. It provides self-adjusting tools.

For users who do not want the activity status bar, this plugin can be skipped.

Watch out — Sandbox testing passed. It was successfully registered in a fresh profile by the harness. No obvious issues found.

The verdict — I would install it because it allows multiple loops in one session and shows a status bar on the chat input, which is convenient for certain loop tasks.

What it is — Provides a VS Code-style workspace keyword search tab in dsh-better-sidebar for DeepSeek Harness.

Who it is for — When you need to search for files in the session workspace of DeepSeek Harness, this plugin provides keyword matches for names and contents that open in the sidebar editor. Not for users without dsh-better-sidebar ≥ 0.4.0, since the plugin registers the search tab through it.

Watch out — Sandbox test passed. Installed successfully in a new profile and registered by harness. No other issues found.

The verdict — I would install it because it allows direct opening of search matches in the built-in editor, but only if you have dsh-better-sidebar ≥ 0.4.0.

What it is — dsh-plannotator enables plan review with anchored annotations and returns structured feedback to the coding agent.

Who it is for — When you use the DeepSeek Harness coding agent to create plans involving multiple independent corrections for architecture migrations, API changes, or security fixes, it provides precise annotation support. If your work mainly involves quick prototyping or has clear execution steps, you can directly let the coding agent write the code.

Watch out — Sandbox test passed: the plugin is registered by Harness in a fresh profile after installation. No obvious issues found.

The verdict — If I were you, I would install it because it requires explicit confirmation when feedback is still pending, under the premise of using the plan mode.

What it is — Import Codex-style sprite-sheet pets and render them as a floating shell.overlay in DeepSeek Harness Web GUI, with pet library and agent-state linkage.

Who it is for — When running agent tasks inside the DeepSeek Harness Web GUI, this plugin fits users who need a persistent bottom-left interactive pet overlay. If using a pure text interface without agent state updates, such a plugin is unnecessary.

Watch out — Requires manually allowing the build script on source install. Static checks indicate it will execute shell commands. No obvious issues found.

The verdict — I would install it because it links directly to agent states for pet animations, but only if the shell command execution risk is acceptable.

What it is — Adds desktop pet as status indicator in DeepSeek Harness.

Who it is for — If you run DeepSeek Harness on Windows 11 and want a desktop pet that switches state to indicate model reasoning or tool use. If you have already obtained pets via hatch-pet or Petdex but do not want them displayed in Harness, you can skip it.

Watch out — Sandbox testing shows it can be installed but the build script for koffi@3.1.5 is blocked by pnpm, preventing Harness from completing registration. Manual allowance of the build is required to fully install.

The verdict — I would install it because it provides visual indication of model state, but I must manually allow the build script first otherwise registration cannot be completed.

What it is — TMCRA-Agent-Memory is a local graph memory solution for DeepSeek Harness and Codex agents. It recalls evidence from owner-global and current-project scopes before each turn and separately records USER and ASSISTANT messages along with provenance to enable visual atlases and cited knowledge pages.

Who it is for — When using DeepSeek Harness or Codex for agent tasks, TMCRA-Agent-Memory is suitable for retaining cross-session project evidence and provenance. It is not suitable for users who do not wish to set up OpenAI-compatible APIs or local generation models, since it requires manual configuration of these to run the memory service.

Watch out — The sandbox test passed: it was installed in a fresh profile and registered in the harness. The static check indicates it will execute shell commands. No obvious issues were found.

The verdict — I would install it because it requires no TMCRA account or production server.

What it is — Zero-dependency tool providing pure function statistical aggregation for numerical data.

Who it is for — When an agent processes numerical data and requires aggregation analysis. Users only needing single-expression evaluation or row/column structure reporting can use calculator or dsh-tool-csv instead.

Watch out — Sandbox test passed in new profile with successful registration. No obvious issues found. Tool execution involves no file reading, no network access, no process creation, and no state saving.

The verdict — I would install it because it provides capabilities that existing toolchains lack.

What it is — dsh-inspect is a DeepSeek Harness plugin that registers checkup, fix, and review tools through the workflow engine to enable an adversarial check-fix-review loop.

Who it is for — When operating in a DeepSeek Harness profile equipped with the workflow engine to trigger checkup fix review tasks via tool descriptions, this plugin provides the integration. If you do not require orchestration using the official workflow engine, then this plugin is not necessary.

Watch out — Sandbox testing passed: installed in a fresh profile and registered successfully by DeepSeek Harness with no runtime exceptions. No obvious issues were found.

The verdict — I would install this plugin because it reuses the official DeepSeek Harness workflow engine for adversarial loops, but only if the profile includes the workflow provider.

What it is — Adds a persistent memory layer to DeepSeek Harness plugins.

Who it is for — When developing agents in DeepSeek Harness that require cross-tool context memory, this plugin is suitable. If you are using other non-DSH platforms, you can skip installing this plugin.

Watch out — Sandbox test passed, no obvious issues found. Historical DSH sessions are not backfilled yet. It only captures new sessions while active.

The verdict — I would install it because it solves persistent memory needs in DSH, but only if nmem CLI is available and for new sessions.

What it is — dsh-spend adds a floating panel to the DeepSeek Harness web UI that displays token usage and estimated spend with per-model, per-day, and per-session statistics.

Who it is for — Developers using DeepSeek Harness for multi-model token tracking would find the panel useful for cost monitoring and billing estimates. Users who do not need UI-based token spend tracking can skip it as it focuses solely on usage metrics rather than general tooling.

Watch out — It passed sandbox testing in a fresh profile and registers successfully with Harness. Static checks note it executes shell commands, but no major issues were found in testing.

The verdict — I would install it if I need to analyze token spends across models and sessions, as it provides direct integration with the UI for real-time stats.