Plugins
Durchsuche, filtere und installiere DeepSeek-Harness-Plugins.
237 Plugins gefunden
skill-bartender
akqwpeter-prog/skill-bartender
Meta-Skill zur Zuordnung von Aufgabe zu Skill mit einer Faulheitsleiter für minimales Laden, einer vom Nutzer editierbaren Routing-Tabelle und einem Installationsablauf mit Quarantäne, SkillSpector-Scan und menschlicher Freigabe.
dsh-plugin-code-review
yytbit/dsh-plugin-code-review
Skill für strukturierte Code-Reviews in DeepSeek Harness — Bugs, Sicherheit, Performance, Stil
dsh-guardian
lonelymoon87/dsh-guardian
Fügt Richtlinienprüfungen für gefährliche Operationen, Ausgabe-Schwärzung und einen Sicherheits-Review-Workflow hinzu.
dsh-mcpguard
chenlaoshiyf/dsh-mcpguard
Durchsucht Skills und MCP-Konfigurationen nach Prompt-Injection, Homoglyphen, verstecktem Unicode, gefährlichen Shell-Befehlen und Zugangsdaten-Lecks.
dsh-fleet-audit
lesliewylie/dsh-fleet-audit
Schreibgeschütztes Audit der Zugangsdaten-Hygiene der Agentenflotte: Berechtigungen von Zugangsdaten-Dateien, eingebettete Zugangsdaten in Git-Remotes (in der Ausgabe maskiert) und Anzahl wörtlich vorkommender Provider-Token; abhängigkeitsfrei und deterministisch.
dsh-dashcam
odysseykernelt/dsh-dashcam
Records what installed plugins actually do at runtime — file access, network endpoints, and commands — in a tamper-evident ledger. It records only and does not block. Its HTTP routes have no DSH access check, so a web page open in your browser could call them.
dsh-workspace (dsh-plugin-airlock)
gabrielbav/dsh-workspace
Puts every plugin, skill and subagent bundle through 6 security scanners (Cisco Skill Scanner, SkillScan, Cisco MCP Scanner, OSV-Scanner, Trivy, Gitleaks) and an optional LLM judge on your own DSH model, in a quarantine with install scripts off, checking for prompt injection, MCP tool poisoning, vulnerable dependencies and leaked secrets; then it installs, asks you, or blocks, and confirms the installed files are the ones it scanned. Works on stock DSH.
dsh-workspace (dsh-code)
gabrielbav/dsh-workspace
Codex cloud–style coding sessions you host yourself: pick a GitHub repo and branch, and each session gets its own checkout and branch, a live preview of the app it builds and an Environment tab, inside a gVisor sandbox (non-root, no capabilities, no Docker socket) whose traffic passes an egress allowlist you extend host by host; your GitHub token never enters it. Runs on your computer or your own VM over SSH.
dsh-workspace (dsh-plugin-aws)
gabrielbav/dsh-workspace
Gives every DSH session AWS's official MCP servers: 59 servers in 14 groups (documentation, IaC, EKS and ECS, Lambda, DynamoDB and Aurora, CloudWatch, cost and billing, Bedrock and more), switched on from a Settings page by group or one at a time. Versions are pinned, servers start read-only where they support it, and secrets stay in the DSH credential store.
dsh-workspace (dsh-connectors)
gabrielbav/dsh-workspace
One Settings page that connects Linear, Jira & Confluence, GitHub, GitLab, Trello, Granola and Azure DevOps to every DSH session, including chats, subagents and automations, with OAuth or a token; add any remote MCP server by URL. Tokens stay in the DSH credential store, and OAuth tokens renew themselves.
dsh-workspace (dsh-plugin-session-delete)
gabrielbav/dsh-workspace
Delete sessions for good: DSH can only archive them, and deleting a workspace leaves its sessions behind. Adds a trash button and menu items to delete a session (with its subagent sessions) or a workspace with all its sessions, and never touches one that is still running or open in another process.
dsh-screensaver
realroc/dsh-screensaver
macOS screen saver for stepping away while AI coding agents keep working: covers the screen with a lock-screen-style cover that a touch of Touch ID dismisses (any key or click brings up the system Touch ID / password dialog), keeps the display and system awake while it is up, can optionally stay awake with the lid closed, and lets themes list live Claude Code, Codex and DeepSeek Harness sessions. Opens with Control+Option+Command+L, the /screensaver command, a screensaver agent tool or after a configurable idle time; also runs standalone with npx. A screen cover, not a security boundary.
dsh-plugin-reverse-proxy
llifco/dsh-plugin-reverse-proxy
Reverse proxy for the DSH Web UI that forwards a LAN port to 127.0.0.1:3080 with Host/Origin rewriting and token-free login. Listens on 127.0.0.1 by default; exposing it to the LAN requires an explicit switch with a security confirmation.
dsh-jev-auto-approve
vultwo/dsh-jev-auto-approve
Reviews DeepSeek Harness approval requests with TypeSafe Jev and auto-approves low-risk, explicitly authorized actions; includes an English settings page for the key and token/cost usage. Each approval request is sent to the third-party TypeSafe Jev service (api.typesafe.ai) with your key, and low-risk ones are approved automatically.
dsh-default-full-permission
18874530790/dsh-default-full-permission
Makes every new and existing session resolve to the danger-full-access sandbox with approval set to never, and pins sessions that still carry a weaker permission.
dsh-lan-access
cy-w-2000/dsh-lan-access
LAN access gateway for the Web UI: serve DSH on localhost or 0.0.0.0 with password protection or an explicit no-password mode plus expiry presets, and open a separate chat-only port that other agents can drive over HTTP with selectable workspace, token or token-free auth, and its own expiry. LAN sharing is opt-in: it rebinds the Web UI port to 0.0.0.0, gives other devices the full DSH interface behind a password (the default), and an explicit no-password mode needs you to confirm a security warning.
dsh-codex-security
lornatang/dsh-codex-security
Agent-native static security review workbench with scan artifacts, finding triage, and SARIF/JSON exports.
dsh-workspace-audit
tomjerrygit/dsh-workspace-audit
Stable local-first workspace health audit for DeepSeek Harness: Git, tests, CI, size, security hotspots, baselines, and bilingual handover reports.
dsh-slowmist-security-guard
wjl906/dsh-slowmist-security-guard
Security guard for DeepSeek Harness (Cordis): reviews every tool call in pre-execute against SlowMist agent-security rules, denying or flagging risky commands, MCP servers and skill loads, external URLs and repositories, untrusted documents, and EVM or Solana addresses; scans tool results for prompt injection after execution and ships a slowmist_scan audit tool plus an injected security policy prompt. Rules follow the open-source SlowMist Agent Security Skill (MIT).
dsh-redact-gateway
luokexiaoguo/dsh-redact-gateway
Replaces identity numbers, phone numbers, bank cards, emails, addresses, customer names and secrets with numbered placeholders before a model request leaves the machine, restores the response, and keeps an encrypted local mapping plus a hash-chained audit log exportable as CSV.
dsh-safety-inspector
ljttt0824/dsh-safety-inspector
Read-only behavior audit for DSH sessions: flags unsafe actions in 8 categories (A–H) with rules for 14 subclasses (the 41-subclass taxonomy file is not shipped, so a built-in 14-subclass fallback is used), anchors every finding to a real event seq, and gives deep links plus the full tool-call / data-access chain of any assistant reply (parameters, results, files read and written, domains, sandbox and approval context). Every finding is drill-downable.
dsh-safebox
jiangzhenguo/dsh-safebox
Adds a DSH Settings panel for write-only local credentials and exposes only secret metadata and references to agents.
mcp-hub
jiale-cheng-ning/mcp-hub
DeepSeek Harness tools to discover MCP configurations across clients, audit static security rules and probe a named server, using a separately installed mcp-hub Rust binary.
dsh-ps-preflight
dingchenhui0618-arch/dsh-ps-preflight
Grades every PowerShell command an agent is about to run before it runs: deletion, overwrite, network download and credential access are four required classes, plus dynamic execution, system and privilege changes. Verdicts are a pure deterministic rule function — no model call, no network, sub-millisecond — with two escalation rules (download+execute, credential+exfiltrate) that jump straight to high. High risk stops the tool call for a second confirmation (in-panel queue or the native approval), deny or timeout means it never runs; medium is logged but never blocked. Every command and its verdict, including the redacted command text, is appended to a local JSONL audit log. Quoted danger strings are treated as data, -WhatIf is respected, -EncodedCommand is decoded and analysed again. Host + client halves, zero runtime dependencies, 55 rules, 136 tests.