Plugins

DeepSeek Harness plugin: web browsing tools (markdown / screenshot / pdf) powered by Cloudflare Browser Run — real headless Chrome with JS rendering, login sessions and WebMCP support.

Harness 事件流审计面板：观察事件类型、分发模式、计数与最近事件，帮助插件作者理解 harness 内部发生了什么

Design I/O for coding agents: controllable UI generation via declarations (spec/domain/craft/design/components/template) and contracts (skill/evaluator). Use for product UI—console, dashboard, agent-ops, CJK-first apps.

Browser Web Speech API voice input: zero server, zero keys, zero model downloads (Edge=Azure, Chrome=Google speech).

Self-editing desktop workbench for DeepSeek Harness: a full dsh web replica with an edit mode in which the agent modifies the workbench (patch layers, plugins, UI) with hot reload.

/repro exports a minimal, secret-scrubbed, replayable problem bundle: the session log, failed commands, and Git diff.

The `dsh pm` plugin manager: multi-source search (awesome list + GitHub + npm), install/remove/update per profile, and a doctor audit of manifests, bundle patches, and version drift.

Adds dangerous-operation policy checks, output redaction, and a security-review workflow.

Adds approval-gated Git status, diff, log, commit, branch, and optional checkpoint tools.

Reusable Markdown prompt profiles for DeepSeek Harness with per-turn model selection, argument substitution, and state restoration.

Turn-scoped “Allow for this task” approvals: automatically allow matching `danger-full-access` escalations only for the current task, then expire.

Agent-tree token budget management.

Microsoft cross-platform sandbox support.

Telemetry backend exporting turns, model steps, and tool calls to yiTrace.

Enforced pain-point gate: after two non-converged experiments it injects the three questions, denies non-investigative tool calls until answered, and blocks same-direction retries.

Feishu (Lark) IM bridge for DeepSeek Harness via `dsh plugin add`; one DM user to one persistent dsh session, with full debugging docs.

Native OS notifications for approval and question/plan prompts: the toast names the workspace, click jumps to it, bilingual (zh-CN/zh-TW/en), with chime.

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.

Render Markdown to a standalone, self-contained HTML page: an `md_html_render` tool that works in a headless profile, plus a web drawer to browse, preview, edit and export local `.md` files; both share one renderer, with no runtime dependencies.

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.

Multi-engine reverse image search aggregator: Google Lens, Baidu, Yandex, TinEye, SauceNAO, IQDB, Ascii2d.

Thin on-demand context retrieval: context_query / context_slice / context_grep tools that read already-persisted history back with replay-safe citations.