Plugins

Independent read-only acceptance layer: spawns a read-only verifier before each top-level turn closes and steers non-pass gaps back into the agent.

Scans skills and MCP configs for prompt injection, homoglyphs, hidden Unicode, dangerous shell, and credential leaks.

Read-only Windows device information tools: one agent tool per Win32 device category (time, system, cpu, memory, disk, gpu, network, battery, processes, usb, audio, printers) via WMI/CIM and Node os collectors.

Read-only agent-fleet credential hygiene audit: credential-file permissions, embedded credentials in git remotes (masked in output), and provider token literal counts; zero-dependency and deterministic.

Per-agent on-demand tool discovery and progressive schema disclosure.

Cited memory over DSH's lossless session log: distilled facts carry `(sessionId, eventRange)` citations that expand back to the exact original log excerpt.

Auto-inject user-chosen skills (e.g. caveman, ponytail) into DeepSeek Harness sessions.every prompt or once at session start.

Telegram control plane for DeepSeek Harness — one group per workspace, one forum topic per session.

Connect your own machines to a DSH Web deployment and run commands on them from the agent.

A composer slime that grows on the output tokens your agent writes, splits into a colony, wears a breed hat from the tool mix, molts on every context compaction, and totals the colony across sessions in a sidebar tab.

Connects personal WeChat direct messages as a dsh-superagent Source, so private-chat text becomes a trigger for the orchestrator, accepted tasks run as work orders, and results are replied to the same chat; group and non-text messages are dropped. Only WeChat users on the superagent policy's allow-list are assessed; the list is empty by default, so every DM is ignored until you add your own ilink_user_id.

Drive DeepSeek Harness from WeChat: scan a QR code to sign in, send instructions in WeChat, and the agent reply lands back in the same chat window. Several agents can share one WeChat account, each with its own model, preset and context, switched from the chat with /use. Only the configured owner WeChat account can drive the agent; other senders are rejected by default and require explicit approval.

Session-level root privilege bridge for DeepSeek Harness: ask mode with per-request web approval, full root mode (polkit password entry in the web UI) and fakeroot mode, with sudo or run0 --empower escalation and a configurable subagent sudo policy. It escalates real commands with sudo or run0: in ask mode each request waits for your approval; in full root mode every bash command of the session runs elevated after one password entry in the web UI. Delegated subagents need per-request approval by default (subagentPolicy: request).

JetBrains IDEA agent preset for DeepSeek Harness that adds an IDEA mode with IDE MCP tools plus projectPath, parameter-name, and tool-name compatibility helpers.

Generates DSH host-side plugin skeletons — tool registration, an approval hook, a dsh.bundle manifest and three test layers — and registers a plugin_kit_new tool so the agent can scaffold a plugin from inside DSH without a shell.

Turn real sessions into regression cases, replay them in isolated git worktrees, and verify tests, changed files, step and token limits. Fully local. Replays run each case's test commands in a git worktree; its HTTP route has no DSH access check.

Before you say yes to an agent, ask how the code works: answers come with an interactive UML diagram (8 types) linked to the text, so pointing at a paragraph highlights it in the diagram and clicking a step jumps to its paragraph. Code quotes are checked against the files.

Batches independent tool calls into one model round-trip: run_tools dispatches many steps concurrently (after-dependencies for ordering), run_parallel fans out sub-agents, so fewer round-trips and lower wall-clock time.

Unread mail widget for DeepSeek Harness: 7-day and 30-day unread sections with senders, a local widget page and two agent tools, IMAP over TLS with no runtime dependencies. The IMAP app password is stored in plain text in ~/.dsh/mail-widget/config.json (mode 600), and the two agent tools return sender and subject lines from your mailbox to the model.

External wake bridge for DSH sessions: an out-of-host watcher posts an event to a loopback endpoint, and a failed delivery records nothing so the caller can retry with the same event_id.

Line-level provenance and a decision ledger for agent edits: which session, model and tool wrote each change, with Markdown/JSON export.

Kaipu Workstation — a cross-cloud, cross-platform console where agents collaborate on one workspace. Any agent runtime that speaks plain HTTP can take a seat, each with its own device credentials: project context is shared, files are kept per seat, and tasks rotate between seats with a trace of who did what. Workstation data lives on the server — this plugin is only the interface, so uninstalling loses no data and reinstalling restores it. The panel renders the persisted workstation state (no live stream yet). Workstation data is stored on the author's hosted server (agent.ak2029.com/kai-pu) unless you override baseUrl in the profile config.

An audit workspace for DeepSeek Harness: a multi-agent consultation venue where each review seat is filled by an agent, every judgment is routed to a named review face, and the whole run is kept as a traceable record. Works out of the box: the service endpoint is built in, a workspace is provisioned automatically on first launch, and review scenes come with their seats and agents listed. The built-in service endpoint is the author's hosted server (agent.ak2029.com/kai-pu): scenes, runs and review records are stored there unless you override baseUrl in the profile config.

ZenTao workbench for DSH: load a ZenTao browse-page URL into a panel, multi-select bugs, tasks and stories, and hand them to the current or a new session with description, recent comments, screenshots as real image attachments and other files saved locally; when the agent is done it writes comments and status transitions (resolve, close, activate, start, finish and so on) back into ZenTao, one item or a batch, from the panel or through five registered agent tools (bug write-back verified end to end, task and story forms share the same code path).