Security & Permissions

What it is — Integrates nine redteam security research modes and fifteen runtime plugins.

Who it is for — Suitable for installation when redteam security research needs mode support and tool plugins. Not needed if you are writing normal business code.

Watch out — Sandbox test passed: successfully registered in a fresh profile. It will execute shell commands. No other obvious issues found.

The verdict — I would install it for authorized redteam security research because it provides structured modes and tools, but only under the premise of authorized use.

What it is — Integrates authorized pentest mode into DeepSeek Harness, with exploration chain, assets and findings displayed via Web view.

Who it is for — When users need to record pentest targets, clues and assets in DeepSeek Harness, this plugin is suitable to install. For users running on Node.js versions below 22.5, this plugin is not suitable because it depends on the sqlite backend.

Watch out — Sandbox test passed: successfully installed in a new profile and registered by Harness. No obvious compatibility issues found. However, runtime requires Node.js >= 22.5 to use the sqlite backend.

The verdict — I would not install it because it records data only within a single session and does not support cross-session continuation.

What it is — helmd is a DeepSeek Harness plugin. It includes a bootstrap tool, a domain router, and 25 reverse-engineering tools.

Who it is for — If you need to integrate security analysis tools across multiple domains, you can use helmd. If you do not want the plugin to execute shell commands, you can skip helmd because it will execute shell commands.

Watch out — Static checks indicate the plugin executes shell commands. Sandbox testing has not been conducted. No obvious issues found.

The verdict — I would not install it because it requires manual subdirectory installation and has not been sandbox tested.

What it is — dsh-auto-review adds a second model as a read-only reviewer subagent to the approval answerer chain in DeepSeek Harness.

Who it is for — When you are handling tool request approvals in DeepSeek Harness, it is suitable to install dsh-auto-review to obtain second model review. If your work does not involve tool request approvals, you do not need to install it.

Watch out — It requires providing an API key or token and may execute shell commands. The sandbox test passed: successfully installed in a fresh profile and registered by harness.

The verdict — I would install it because it can feed the reason back to the calling model when denied, avoiding blind retries.

What it is — Inserts declarative permission rules in the DeepSeek Harness tools/pre-execute stage.

Who it is for — When using any model with DeepSeek Harness for precise tool call control, the plugin is suitable. For users who only need the default sandbox preset, they can skip this plugin.

Watch out — Sandbox test passed: successfully installed in a new profile and registered into the profile. The static check indicates it will execute shell commands. No other obvious issues found.

The verdict — I would install it because it provides full session-log audit and supports hot reload.

What it is — Detects prompt injection, redacts Chinese PII, and performs local configuration audit.

Who it is for — Users needing security checks when using DeepSeek Harness are suitable for installing this plugin. If you focus on tool calls rather than security review, you don't need this plugin.

Watch out — Sandbox test passes: installed in fresh profile and registered by Harness. The plugin only performs reads, with no write or delete risk. No obvious pitfalls found.

The verdict — I would install it because it audits configuration read-only and provides reproducible reports, assuming use of supported dsh-tools version.

What it is — dsh-approval-gate provides automated approval gate for DeepSeek Harness write commands.

Who it is for — Users executing write commands in DeepSeek Harness are suitable to install it because it uses Flash model to auto-classify risks and approve. Users not working in the DeepSeek Harness environment can skip installing it because it requires DSH profile for registration.

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

The verdict — I would install it because it supports session-scoped snapshots and one-click revert for file diffs.

What it is — Adds secure remote access gate to DeepSeek Harness Web UI.

Who it is for — For users accessing DeepSeek Harness Web UI remotely via tunnels, this plugin provides login gate, MFA/TOTP, signed session cookies, optional roles, and in-browser workspace selection. For users requiring multi-account isolation, this plugin is unsuitable as DSH core is single-tenant.

Watch out — Requires API key or token. Sandbox test passed in new profile and registered by harness. Config changes require restarting dsh web.

The verdict — I would install it because it supports in-browser workspace selection and file previews, but only if running a local DeepSeek Harness instance.

What it is — Adds a server-grade multi-tenant plugin to DeepSeek Harness for remote access and auto HTTPS.

Who it is for — When sharing DeepSeek Harness on a server for team use requiring subuser permissions and token/daily quotas. If your usage is purely local single-machine without multi-user access and permissions management, you can skip it.

Watch out — Installation requires no extra configuration. Sandbox testing passed in a new profile. No obvious issues found.

The verdict — I would install it because it enables subuser permissions and token/daily quotas for team sharing on servers, but only if not using it locally.

What it is — Integrate 24 git-diff audit rules for commit-time auditing.

Who it is for — Users of existing DSH who want their AI coding agents to follow rules.

If you want a full commercial agent platform, you don't need to install this because it's a governance layer, not the platform itself.

Watch out — Sandbox testing not performed, no obvious issues found.

Static check shows it will execute shell commands.

The verdict — I would not install it because it is a governance layer plugin, not a complete platform.

What it is — DeepSeek Harness plugin enabling remote web launch with username/password authentication.

Who it is for — If you need to securely access the DeepSeek Harness web interface from your local network, this plugin enables username/password login. If you only use the command-line interface of DeepSeek Harness without needing the web UI, you do not need this plugin.

Watch out — Requires providing an API key or token. Sandbox testing passed: successfully installed in a fresh profile and registered with harness. No obvious issues found.

The verdict — I would install it because it allows secure remote access to the DeepSeek Harness web interface with authentication.

What it is — DeepSeek Harness penetration testing plugin with Root-Orchestrator architecture.

Who it is for — In DSH Root sessions that require pentester persona preset. Users not using MCP transport can skip this plugin.

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

The verdict — I would install it because it achieves native integration with DeepSeek Harness Skill Runtime V4.

What it is — dsh-defend detects prompt-injection, jailbreak, and secret-leak patterns in DeepSeek Harness agent pre-step and tool execution seams, implementing allow/ask/block decisions with audit events.

Who it is for — In DeepSeek Harness agent pre-step processing user messages, users who need prompt-injection protection can install this plugin. Users who do not use any tools need not install it, because the plugin targets tool pre-execute and post-execute interception points.

Watch out — Sandbox testing passes in a fresh profile and registers successfully. Static checks note that it executes shell commands. No other issues found.

The verdict — I would install it because it provides audit events across multiple interception points and requires no model-specific adjustments.

What it is — DeepSeek Harness plugin that inserts Auto permission preset, letting classifier auto-approve routine sandbox escalations once.

Who it is for — Suitable for developers handling long tasks and dependency installs who want to reduce manual interruptions with a classifier model. Not suitable for users who want to avoid any human approval, because dangerous operations still require manual review.

Watch out — Tested successfully: installed in fresh profile and registered with harness. Classification calls have timeout that falls back to manual. No obvious issues found.

The verdict — I would install it because it reduces interruptions for routine tasks, but only if you configure a classifier model to avoid timeouts.

What it is — This is a skill pack providing eight executable security audit methodologies for DeepSeek Harness.

Who it is for — In scenarios requiring whole-repo security audits using DeepSeek Harness agent models lacking security methodology skills, this plugin is suitable. Users who only want to execute scans using tool plugins should not install it, as this pack focuses on methodology teaching rather than scan execution.

Watch out — Sandbox test passed, it gets registered after installation in new profile. Static checks indicate execution of shell commands, no other issues found.

The verdict — I would install it because it combines security audit methodologies with pre-install scanning.

What it is — dsh-auth-gate adds a login gate to the DeepSeek Harness web surface using password or shared token authentication to protect agents, chat sessions and LLM credentials with a user-management CLI.

Who it is for — When deploying dsh web publicly to protect agents, chat sessions, and LLM credentials from unauthenticated access using the recommended password mode, this plugin fits. Those without the web profile may skip it since it only protects the web surface.

Watch out — Sandbox test passed with installation and registration in a new profile. Static checks show execution of shell commands and requirement to provide an API key or token. Configuration requires patching cordis.patch.yml and restarting the service.

The verdict — I would install it because it offers secure default protection for the web surface along with the user management CLI, but only if the deployment uses HTTPS for secure cookies and the server is properly secured.

What it is — A local security audit plugin integrated with DeepSeek Harness, generating redacted risk reports.

Who it is for — When using the headless profile of DeepSeek Harness for auditing local configuration, plugin origins, and sessions, this plugin is suitable. Those primarily doing remote model operations without local file exposure checks do not need it.

Watch out — Sandbox test passed: installed into a new profile and successfully registered by harness. No obvious issues found.

The verdict — I will install it because it supports report actions for risk summarization in local environments.

What it is — DeepSeek Harness Web authentication gateway plugin, supporting password and TOTP 2FA authentication.

Who it is for — In scenarios requiring remote access to DeepSeek Harness Web, people lacking authentication layer when using DeepSeek Harness for web work are suitable to install it. People not using external ports are not suitable to install it, because the plugin uses external ports for authentication.

Watch out — Sandbox test passed: installed in a new profile and registered into the profile by harness. No obvious pitfalls found.

The verdict — I would install it, because it complements the authentication layer for dsh web.

What it is — Integrates white-box causal adjudication engine to adjudicate tool calls before execution in DeepSeek Harness.

Who it is for — When using DeepSeek model for multi-turn tool call agent tasks, this plugin applies to ensure causal boundary safety. For users who only need fixed script execution, this plugin does not apply.

Watch out — Successfully registered and loaded after installation in a new profile. Static check shows it will execute shell commands. No obvious issues found.

The verdict — I will install it because it offers six white-box self-check tools that run locally with zero API cost.

What it is — Migrates Claude Code configurations into DeepSeek Harness, previewing instructions, skills, and permission rules before application.

Who it is for — In scenarios where it is necessary to migrate Claude Code instructions, skills, and MCP servers into DeepSeek Harness, this plugin is suitable to install. If the configurations have already been completed in DeepSeek Harness, there is no need to install this plugin.

Watch out — No obvious issues were found. The sandbox test passed by installing in a fresh profile and registering it with harness.

The verdict — I would install it because it can safely migrate configurations without directly modifying existing files.

What it is — upstream-radar tracks exact dependency paths of DSH plugins, OSV vulnerability signals, and npm release changes, then routes project evidence to a DSH Agent.

Who it is for — When a DSH profile contains multiple third-party bundles at once, monitoring dependency security becomes necessary. Profiles that use only one built-in plugin or zero third-party bundles do not need this plugin.

Watch out — Manually allow the build script when installing from source. It executes shell commands. Sandbox testing confirmed successful registration in a fresh profile; no obvious issues found.

The verdict — I would install it because it supplies dependency-path analysis for profiles that contain more than one bundle.

What it is — The dsh-auto-approval-llm plugin provides LLM-assisted auto approval for the DeepSeek Harness Auto permission preset.

Who it is for — When processing approval/request in a DeepSeek Harness Auto session, use this plugin. It is not suitable for use in Android browser environments because the Auto permission preset does not support Android native environments.

Watch out — Sandbox testing passed: it installs and registers into the profile in a fresh profile. No obvious issues found.

The verdict — I would install it because it supports declarative rules.