Plugins

Turn event notifications: six event classes (completed, error, interrupted, approval requested, AI question, limit reached) delivered via sound, system notification, in-page banner, webhook, IM and host desktop-notification channels; when several browser windows are open, only one rings.

Ships a skill about handing work to subagents instead of doing it inline - which tasks are worth fanning out, how many to run at once, how to send them in parallel, and how to check what they returned instead of trusting the summary.

Speaks or plays a sound at the end of every turn and plays a failure cue when a tool call or a turn errors. Ships 20 built-in effects (10 alert chimes + 10 nature sounds) and defaults to effect mode, so it needs no API key and no audio files; an optional Volcengine voice-clone route generates the three announcement clips in one click. Plays through winmm/waveOut as-is, never changing the system volume or mute state. Windows only.

Agent screen capture on macOS and Windows: one tool captures the screen and returns the image itself, so the model sees it without a second call. On Windows a whole burst runs in one engine call; on macOS a resident helper drops a region capture to 13ms and a change check to 23ms. A second tool, macOS only, reports whether Screen Recording is granted and opens the settings pane that fixes it.

Memory archive and prompt viewer for DeepSeek Harness: read back what compaction folded out of context - browse past sessions or Tavern playthrough archives (summary, original text, state) with full-text search - and inspect the prompt each turn actually sent, listing every assembled system section with its order, character count, offset, hash and mutability, plus an assembly map, registry export, and honest diagnostics when a section is empty or its log cannot be read. Optionally provisions a roleplay agent preset.

On-demand memory for DeepSeek Harness: session logs are distilled into layered markdown (a summary, an index and per-topic files), and a mem_query tool returns only the matched entries rather than the whole index. Nothing is injected per turn, there are no dependencies, and every entry keeps a source pointer back to the conversation it came from.

Inject steering or correction into a running session while the model is thinking, or cancel the current turn, from a floating browser panel.

Session usage statistics panel: token and request stats at day/hour/minute granularity, an activity heatmap, bottom-bar session info takeover with per-turn cost chips, and cost estimation under multi-condition time-segmented pricing. Re-implemented from HaoyueQin/dsh-usage-statistics-panel with extensions.

System-audio visualizer for the turn navigator rail: the rail reacts in real time to whatever the default playback device is rendering, with four motion modes (spectrum, pulse, sway, travelling wave), a 16-band logarithmic spectrum from a 2048-point FFT on the host, and live sensitivity, strength, bass-punch and wave-speed controls. Audio is captured host-side through WASAPI loopback, so there is no permission prompt, no microphone and no screen capture. The browser half adds one attribute, one stylesheet, some CSS variables and a sidebar control row, and never patches the official component; switching it off removes the attribute, the stylesheet and the variables, returning the rail to its default styling. Windows only.

Per-turn token usage, request composition and cost breakdown panel for the DeepSeek Harness Web GUI.

Turn approved tracker tickets into review-ready pull requests with DeepSeek Harness.

In-process DeepSeek Harness subagent provider that seeds a child with only the last N completed parent turns (fork-lite), leaving official fork untouched.

Engineering work-tools for DeepSeek Harness: registers two model-facing Host Tools — a scoped `git` runner, and a package-manager runner (`dev`) that detects npm/pnpm/yarn/bun from the workspace lockfile and picks the right verb per manager. Both run through dsh's sandboxed shell and return structured stdout/stderr/exitCode, so the agent gets focused, reviewable results instead of raw bash, and adding another tool is a copy-paste of one register block.

DSH frontend design generation skill: turns a design request into a polished, self-contained HTML/SVG artifact you can open by double-clicking (poster, infographic, landing page, chart, component mockup). Adapted from Anthropic canvas-design / web-artifac

Byte-faithful text editing tools (edit_text / write_text) for DeepSeek Harness. They fix what the built-in write/edit cannot do: both drop a UTF-8 BOM, neither restores the file's own line-ending style (writing LF content turns a CRLF file into an LF file

DSH plugin: resumes a turn that ended successfully with reasoning only, no text and no tool call, by sending one bounded follow-up

OpenReelbench, the open video creative bench for DeepSeek Harness: a gated four-stage pipeline (brief -> script -> assets -> compose) that turns one sentence into a narrated explainer video; generation via ComfyUI workflows, assembly with FFmpeg. / DSH 开源

Turn the DeepSeek Harness Web UI into a real desktop application from inside DSH: one plugin that provisions a portable launcher (native WebView2 window, tray residency, silent backend start, readiness gating, debug console) plus desktop/start-menu shortc

Review agent-produced file changes (line-level diff + undo) as a turn-tail row and dsh-better-sidebar tab. Fork of dsh-file-review-tab@0.4.1 patched for DeepSeek Harness 0.1.2-alpha.1.

Per-turn and whole-session DeepSeek API cost estimates for DeepSeek Harness Web.

DSH Agent 文件修改时间机器：session baseline、turn 级修改记录、安全恢复与 conflict 检测

DeepSeek Harness (DSH) web plugin: per-turn response window. Each turn's tool calls collapse into ONE always-expanded bounded slide (configurable N-line window, inner scroll, nothing hidden), and long assistant markdown gets a bounded scroll window too —

Keeps an unfinished task list visible across turns. DSH clears the todos projection at the start of every turn and expects the model to rewrite the whole list, so when a turn is interrupted the plan is durably in the session log but permanently gone from the UI — and todo_write is write-only, so neither the user nor the model can read it back. This plugin registers one ordinary session/event listener, remembers the last list per session, and when a turn/start arrives whose remembered list still has unfinished work, re-appends that list so the projection stops reporting an empty plan. No fork of tool-todo and no waterfall: a plugin cannot replace the projection (the registry throws on a stateVersion mismatch), so appending the event the fold reacts to is the reachable seam. Deliberately narrow — only unfinished lists, at most one event per turn start, fails open. Verified by a controlled A/B on real sessions: identical runs differing only in whether the plugin was installed, folded through the real projection definition. Zero dependencies.

Keeps one reasoning-only turn from making an entire session unusable. When a turn produces no visible text and no tool call, the assistant message is persisted empty, the gateway then rejects every later request in that session with 'content or tool_calls must be set', and the session is dead with no way back through chat. This plugin registers a single llm/stream waterfall listener and, only when the turn carried nothing visible, injects a small text block immediately before the terminal finish, so the persisted message is never empty. Ships the fixture DSH's own mock cannot express (a strictly reasoning-only SSE server, since llm-mock-server forbids an empty successText and always appends text after reasoning), plus an end-to-end reproduction that replays a real persisted session through DSH's own serializeMessages. Preventive, not a repair. Zero dependencies, one file, no process or filesystem access.