Plugins

Codex-style desktop pet: a floating animated sprite in the corner that mirrors the agent's running state (working, waiting, failed, done).

Short-video sidebar: native player, series navigation, precise history replay.

Plugin store: npm authoritative catalog + awesome curated list (550+ plugins, 11 categories), dsh-field quality verification, official `dsh plugin add/remove` one-click install, sidebar + settings entry.

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.

Per-model token cost tracking with configurable cache-hit/miss, output and peak-window pricing, a live session cost bar, and unconfigured-model flags.

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

Native interactive Git diff review for DeepSeek Harness with structured annotations sent back to the current Agent session.

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.

Generate Functions, MCP tools, workflow Skills, and offline test packages from user-authorized source code.

Deterministic keyword routing, allowlisted model switching, and an isolated image_gen channel for DeepSeek Harness.

Reuses the Codex CLI ChatGPT login as an `openai-codex` LLM route and adds GPT Auth controls to DSH Web settings.

Adaptive model routing: per-request complexity classification with automatic provider routing.

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

Connect a Feishu bot to DeepSeek Harness by scanning a QR code.

Two-way WeChat (iLink) bridge: turn-end and approval-request push, in-chat approve/reject and message injection, persistent dedup and convergent long-reply chunking; channel layer extensible to other IMs.

Two-level model circuit breaker with failover: trip a model or a whole provider after repeated request failures and route the next request to a configured fallback.

Model-based permission approval: an approval-request answerer backed by a separate reviewer model.

Engineering-discipline guard: grill the requirements before the first edit, enforce red/green test evidence gates, and audit the delivery with a forked adversary (grill-requirements skill + tool-policy gates).

SHOPLINE AI Toolkit: bridges the official SHOPLINE Developer MCP server and ships seven SHOPLINE agent skills (Admin REST, GraphQL, OAuth, webhooks, Sline) — the SHOPLINE counterpart of the Shopify AI Toolkit.

Dual knowledge-graph RAG for Markdown notes: prerequisite + similarity graphs (P/S-Agent traversal), paragraph chunking, n-gram/embedding hybrid search, and an explicit weight feedback loop — via MCP.

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.