プラグイン

Left sidebar panel list for DeepSeek Harness: long-press drag to reorder panel icons, one arrow to fold them all, whole-row height caps, a pinned-session section that keeps its live running / waiting status, draggable conversation view tabs, archived Sess

深鲸记忆：跨会话热记忆（每轮注入 MEMORY/USER + memory 工具读写）+ 记忆体检、瘦身归档、记忆包、诊断报告、文档只读。本地明文、零依赖、不上传数据。

A DSH companion with a macOS native pet and a Windows/Linux Electron companion. Shows the DSH account balance on the pet even when idle, plus color-coded task cards, progress and notices.

Hides the dsh-pet desktop pet while a game or fullscreen app is in the foreground, then brings it back. Decides from three signals — a captionless window covering the whole monitor, the foreground process living in a game directory, and the system's full-screen notification state — and requires two consistent polls before acting, so alt-tabbing does not make the pet flicker. It only toggles window visibility; dsh-pet's configuration and install directory are never touched. A tray menu offers pause, restore now, and reload config. Windows only.

Floating desktop companion for DSH: a draggable character on the overlay — tap to charge energy, blinking and emotes, naps, synthesized WebAudio sounds and particle effects. Bring your own: any photo (pet, portrait, anime art) becomes a 2.5D cutout with hand-marked eyes, and 3D models (GLB, VRM, STL, OBJ, PLY) render as a real three.js turntable. Resize 30%-300%; data stays local.

Six-yao (I Ching) divination for DeepSeek Harness — xiantian bagua backdrop, a Zhouyi opening verse, manual line entry or a simulated three-coin toss, najia charting (six relatives, six gods, void branches, hidden spirits, nayin, seasonal strength, transforming lines), an AI interpretation that cites its classical source for every statement, and a one-hexagram-one-question follow-up dialog. Zero npm dependencies; every request goes to the local DSH host, with no third-party network access.

Adds a token-burn leveling widget to DSH with 14 DeepSeek-whale levels, daily quests, streaks and achievements.

Remote workspaces for DeepSeek Harness: pick an SSH host in the sidebar, mount a server directory over SFTP as a workspace, then read, write and edit those files with the normal file tools. Remote watch is polled so the file tree refreshes, image/binary previews work through byte-range reads, and remote_exec, remote_search and remote_git give agents the remote shell, search and git queries. An opt-in patch routes the agent's own bash/pwsh commands to the server. Every remote write and command follows the session sandbox mode and approval policy, and hosts are managed by the plugin itself (password, key or SSH agent).

Session password lock plus an encrypted directory vault for DeepSeek Harness.

Exports selected DSH plugins as one pasteable share code, applies the delta on another machine, and reports which credentials still need configuring.

A local workbench with model metrics, cost estimates, prompt presets, and optional request auditing for DeepSeek Harness.

Plays a short bell the moment DeepSeek Harness asks for tool approval or a question, so a stalled turn gets noticed.

Bark push notifications for DeepSeek Harness: turn completion, waiting prompts and approvals, plus opt-in background-job cards. Devices and card settings are edited live, and each push checks the per-device result returned by the Bark server and retries transient failures.

Desktop notifications and an app-icon unread badge for DeepSeek Harness. A banner fires when a turn finishes, when an approval, question or plan confirmation is waiting, and when a turn fails; the badge shows how many sessions need you and clears once viewed, and clicking a banner switches to that session. Follows the DSH interface language, subagent sessions stay quiet by default, titles and body length are configurable, and a loopback HTTP interface lets scripts and other tools post notifications.

DSH secretary: triages your unread GitHub notifications by their real state every 15 minutes (closing what is already resolved), remembers dated things you tell it (bills, renewals, document expiries, birthdays) and reminds you when due, and watches your inbox read-only; secrets live in the system keychain only.

Turn timer for DSH: reports how long the current turn has been running, injects wrap-up prompts on a /set-time schedule so the agent starts finishing instead of running on, and shows run-time cards above the composer.

Sync DSH user-level assets (skills, plugin manifest, config, memory) across machines with one-click restore, automatic plugin reinstall and secret scanning.

Dual-host coordination for DeepSeek Harness: hand a task off in local chat, run bounded jobs in a cloud worker, then review results and approve follow-up steps in one workbench. Includes protected six-page onboarding, per-task model selection, a task approval ledger and stepwise controlled delivery.

Step-by-step workflow map for DeepSeek Harness: folds session events into a step list with per-step status, a percentage whose source (todo table, self-report, time baseline or plan ordinal) is labelled, tool and file-type interfaces either side of each unit, a detail page, and hot injection that queues a prompt for the next step with revoke before it is claimed.

Progressive disclosure agent preset and call directory tool that starts from a minimal baseline and unlocks tools, skills, and MCP servers on demand.

Provides three skills for the Let's Build Robots SO-101 course: yanshifu reads the learner's project files to assess progress and suggest next steps; ask answers questions using four bundled guides; update checks for and installs updates to those guides.

Codex-compatible Skill Studio with a left-sidebar page for browsing, creating, editing, enabling, disabling, and invoking SKILL.md skills; bridges ~/.codex/skills and project .codex/skills into DSH and supports $skill-name invocation in chat.

Turns a YouTube link into study notes: pulls the captions the video already publishes, optionally reads on-screen content with a vision model, then asks for your own notes and merges everything into one document. Requires python3, ffmpeg and yt-dlp on PATH (the skill stops with an install hint when they are missing).

Read PDF, DOCX, XLSX, PPTX and text documents as agent tools, and OCR scanned PDFs or images locally with tesseract.js. Exposes doc_read, pdf_read, xlsx_read, doc_meta and doc_ocr.