Plugins

Plugin governance: new/updated plugins stage in a candidate area first, then hot-mount after startup — success promotes, failure auto-disables, repeated failures freeze, one-click safe mode, sidebar diagnostics panel (npm: `bsfeng-dsh-guardian`).

Browser notifications and a beep when a session turn ends, the agent asks a question or approval is pending; clicking the notification jumps to the session; remote trigger endpoint and SSE channel (npm: `bsfeng-dsh-notify`).

Codex-style review of session file changes: per-turn cards (optionally taking over the shell edited cards), a review tab with syntax-highlighted diffs, real line numbers, open-at-line and a draggable split, light/dark presets that follow the DSH theme, typography that follows DSH and installed skins (or any font on the machine, with a px stepper), and one-click per-turn revert — powered by the official session channel.

Identity-gated Tailscale access to the Web GUI with dsh still bound to loopback: a 127.0.0.1 proxy published by `tailscale serve` admits a request only when the Tailscale login it stamps is allowed, with an optional per-device allowlist, so the privileged /api methods behind the loopback check — Settings, Models, Plugins — answer without dsh being told to trust a remote host.

Archived-conversation manager for DeepSeek Harness: view archived conversations grouped by workspace, restore them, and reveal their transcript folders.

Rewrites the Chinese copy of a market-installed plugin into English using a model already configured in the harness — automatically when the plugin is installed, then live-reloaded. Each file is translated in one whole-file pass, validated by a real parser where one exists (node --check for JavaScript, JSON and YAML parsers otherwise) and handed back to the model with the error for up to three repair attempts; files above 48 KB, or that never parse, fall back to a line-range pass that checks every returned line against the original for structure. Trigger phrases and locale tables can optionally be preserved rather than translated, and the untouched original is kept beside the file as a .zh.bak.

Deep research as an agent mode: one reviewed loop that plans, researches in adaptive parallel rounds driven by the gaps the researchers themselves declare, synthesises a cited report and then reviews it adversarially, instead of a script the model rewrites per call. Questions the round budget never reached are named in the report rather than dropped.

Config-driven lifecycle hooks, including context injection: declare event, handler and outcome in YAML, where the handler is an in-harness tool (MCP included), a shell command or an HTTP endpoint, and the outcome is model-visible context, a denied tool call, or fire-and-forget.

Points web_search at a self-hosted SearXNG instance instead of the native route: pages until maxResults is satisfied, and reports degraded coverage when engines fail so a thin result set is not read as a thin web.

A guarded local fetch provider for the ctx.web seam — the backend dsh does not ship for its own web_fetch tool. Every redirect hop is re-resolved and checked against the non-routable ranges, with a streamed size cap and a content-type allowlist.

Progressive tool disclosure: a deferred tool group costs one catalog line instead of its full schemas until the model loads it with tool_search, with one switch per group in Settings and each group measured against the live registry.

A thinking level stored per provider and model, so reasoning effort outlives the session it was set in and also reaches subagents, which dsh never asks. The in-session picker still wins, and a model with no row keeps its provider default.

Per provider/model generation concurrency, with a settings card: a cap on how many sessions may stream at once, enforced at the stream and again at subagent spawn, where a fan-out past the cap waits in a FIFO queue instead of failing.

After each completed agent turn, a bounded auxiliary LLM call writes one suggested next prompt into the session log; the web composer renders it as ghost placeholder text, adopted via Tab.

Maps locally analyzed system audio to music-reactive conversation effects, rhythm-game judgements, score and combo feedback, and optional RGB spectrum bars in DSH Web.

Multi-lens structural analysis skill: adversarial-first analysis across 16 lenses in 4 categories with 10 structural tools, confidence calibration, and fact-bound layered output.

Off-peak message scheduler: turn on valley mode and the composer queues your messages until DeepSeek off-peak hours (half price), then auto-injects them into the session and gets replies. Native-look composer, hotkeys (Enter queue / ⌘+Enter send now / Esc exit), per-session toggle persisted across restarts.

Windows notifications, zero dependencies.

UI enhancement suite for DeepSeek Harness: one-click workspace open (copy path, reveal in Finder, open in an auto-discovered editor), DeepSeek-web-style user-message nav rail, session to-do marks, flat-list polish, and a hard-restart button.

Runtime extension suite for DeepSeek Harness: cross-session mailbox with wake cold-start, agent teams (captain + members + dependency task board + team_wait), subagent spawn policy with escalation approval, task-aware routing preset, plugin market, skill manager and runtime injector.

Load Claude Code assets into DSH at runtime — skills, slash commands, rules, permission rules (deny/ask/allow), subagents and hooks from .claude/ plus Claude Code plugins (plugin.json / marketplace / skills / agents / MCP), via a memory-IR parse layer that keeps .claude as the source of truth with no writes back. Six installable packages: loader, skills, permissions, agents, hooks, mcp.

IM channel bridge: WeChat (ilinkai) / QQ / Feishu with proactive push — wake the channel bot from scheduled tasks and deliver AI replies to your phone.

GitHub Device Flow connect plus an in-conversation PR workflow: a status bar above the composer for create / AI-review / merge, and github_* tools for search, issue and PR read and write.

Virtual product team agent preset: user-led conversation walks you through the full 12-step pipeline from idea to shipped, promoted and operated product, with staged gates and per-stage skills.