Plugins

Bidirectional GitHub repository sync for DSH: git sync with conflict rollback and pre-push checks, Actions/CI monitoring with per-step run detail, repository add/remove/rename, a web panel with local and remote file browsers, scheduled automation, and six agent tools.

Redundant branch cleaner: finds branches whose work already landed in the mainline — merged, squash-merged, patch-equivalent after rebase or cherry-pick, upstream-gone, or PR-merged — and deletes them when asked. It only reports by default, never switches branches or rebases, and backs every deleted local branch up under refs/backup/branch-cleaner/ (deleting remote branches has no backup).

Registers native doubao_generate_image and doubao_image_status Agent tools to automate Doubao Web (doubao.com) via Tabbit Browser, automatically extracting watermark-free 2.7K wm0 original images into local files and multimodal chat attachments.

Self-maintained DSH web task board with a human acceptance gate, card kinds and editable notes on executed cards.

Evidence-first reuse skill for DeepSeek Harness: before implementing, porting, choosing dependencies, or troubleshooting a tool, it searches the local project and machine for existing solutions in layers, judges reuse on checkable evidence, refuses reuse when essential requirements or risks are unmet, and reports what could not be verified. Ships as a DSH bundle that registers the skill on ctx.skills.

Bundles the kb-curator skill, which keeps a local three-layer knowledge base and refuses to index anything the user has not approved. The L2 retrieval layer needs @kenz1117/dsh-engram installed separately, and the engine runs on Windows only.

Reads and writes WPS 365 cloud documents (smart documents and .docx/.xlsx/.pptx) through the locally installed wps365-cli, with conflict checks before writing; requires a WPS 365 enterprise account.

GPT-SoVITS voice output for DSH: a speak tool and /voice command, a read-aloud speaker button under each reply, and a settings tab to switch voices, tune synthesis, manage the local service and one-click install a voice pack. Windows only; needs a local GPT-SoVITS checkout with trained weights and its Python environment.

Meeting notes for dsh: one meeting per conversation, transcribed and speaker-separated on the machine itself, with points, assignments, handwritten notes, reviews and monthly scores in the same panel. Exports SRT/Markdown/Word and packs a meeting into a portable zip another dsh can import; confirmed assignments go to DingTalk/WeCom in one click, following your robot settings. Recognition models are swappable (a stronger local model can be imported from a file), and cloud recognition is off by default with no request made before consent. Word export needs Python and python-docx; the first run downloads a recognition model (28 or 77 MB).

Local offline prosody analysis for DeepSeek Harness: reports how something was said — speed, pitch, pauses, energy — alongside the transcript.

Registers two tools for DeepSeek Harness. `doc_ocr` turns scanned PDFs and images into structured text (per-page text, financial tables with a balance-sheet tie-out check, contract-clause hits) through a local RapidOCR engine or a Zhipu GLM-OCR cloud engine, and `doc_preprocess` turns a directory of raw files into an Obsidian-ready knowledge base through a seven-stage pipeline with backend routing and sha256-based incremental skipping. Numbers that fail the tie-out are reported, not invented. The local engine needs Python 3 with rapidocr-onnxruntime, pypdfium2 and Pillow (pip install -r requirements.txt); the GLM engine needs ZHIPU_API_KEY.

Reads a local video or page link with ffmpeg and Paraformer ASR, returning a timestamped transcript plus key-frame images the agent can open; finds candidate videos by keyword and installs its own runtime dependencies on first use. Requires Python 3.8+ on PATH; video_discover also needs a one-time douyin/xiaohongshu login.

Structured image recognition: local sharp measurements plus a cloud vision model return description, OCR with coordinates, object locations and fine-grained attributes in one call; a fingerprint cache makes repeated images return in milliseconds without spending model quota.

All-media (video/audio/image) sibling of liustack/modlens: turns local media into structured evidence (timeline / transcript / layout regions) injected into the conversation so text-only models can cite concrete frames, seconds and speakers — Evidence, not imagination. One failover chain per media, recorded hop-by-hop in meta.attempts (a fallback is never silent): host-registered multimodal models first, then a configured OpenAI-compatible endpoint (comma-separated multi-key rotation), then local ffmpeg frame-sampling / a local ASR CLI, degrading to metadata-only as a last resort. Absorbs modlens battle-tested rules — cooldown keyed only on auth/quota failures and softened to chain-tail (never dropped), redact-before-clip error text, magic-byte sniff that never trusts file extensions. Optional vision-bridge registers a media-lens wrapper into the host model selector so a pure-text model gains image/audio/video sight. Registering / authenticating multimodal models and true inline audio-video upload lean on two companion plugins, declared as optional peers (absence auto-degrades, never errors): dsh-thinking-levels fronts the model config / registration surface, dsh-llm-openai-completions takes over the OpenAI-compatible transport that serializes file blocks into video_url / input_audio. Privacy: media goes to remote providers by default; strictLocalOnly keeps content on-machine. Prerequisites not shipped with the package: ffmpeg/ffprobe on PATH (frame sampling / reencode / duration probe; degrades to metadata when missing) and a user-configured local ASR CLI for the audio fallback. Hosts 0.1.7 / 0.2.0 / 0.2.1 (6 rc, exact engines.dsh enum). npm: dsh-media-lens.

Lets the model actually see a video: local ffmpeg samples timestamped stills and faster-whisper transcribes the audio track, so the agent reads real pixels frame by frame instead of trusting a text summary. Fully local, nothing uploaded; reports honestly when the audio has no speech. Windows only; needs Python 3.10+ and ffmpeg installed, and the first video_eyes_setup run downloads about 700 MB (faster-whisper and its model).

Registers one `valuation_service` tool that reads a workpaper tree of 01/02/04 JSON and returns one unified schema: six ratio groups, forecast ratios, DCF input slots with explicit missing markers, comparable PS/PE and IRR. Deterministic and local: missing WACC, growth, capex, D&A or working-capital inputs stay blank instead of being guessed, and assumption differences are flagged for review.

Requires Claude Code installed and signed in on this machine. Delegates coding tasks to the local Claude Code agent through the Claude Agent SDK, reports which files changed, supports one-command revert, and adds a Claude Mode toggle to the composer.

Run Python on the local NVIDIA CUDA GPU from DeepSeek Harness: device status, foreground and detached jobs, durable logs, and a measured peak-VRAM figure per run.

Requires Python 3.12+; Xiaohongshu reading needs the xiaohongshu-mcp backend and a QR sign-in (ffmpeg/OCR optional). Batch-read Xiaohongshu notes and WeChat public-account articles: per-note body, image text via local OCR, video frame text and comment samples, structured JSON/Markdown output, and the original source URL on every result. Read-only: it never publishes, likes, follows or comments.

Local 3D studio orchestration: detects Blender, COLMAP and ComfyUI on every run and picks the route they support, and also ships a Three.js fallback that generates models from primitive specs, edits or scales GLB files, writes an offline viewer page and renders multi-view PNGs without Blender.

Run the local Cursor CLI agent on a coding task in a workspace and continue the run by chat id.

Read Bilibili and Douyin from a session: WBI-signed Bilibili API calls behind one shared pacer (4 req/s, per-host cooldown after a 412) for video, feed, search, author, comments, danmaku and subtitles, plus Douyin hot list, video, feed and search rendered in a locally launched Edge over CDP.

System One decision runtime for dsh. Ships fast non-generative tools (choice, score, screen, gate) on a local or hosted System One gateway, and — without the model calling anything — prunes oversized tool output segment by segment, refuses repeated identical calls once System One confirms stagnation, verifies completion claims against real tool evidence, and corrects failing retries. Published on npm as dsh-harness-reflex (plugin) and harness-reflex (core, also a stdio MCP server).

Built for DSH 0.2.1-rc.2 (engines.dsh pins it exactly). Technical support for Dragonfly 3D World using 2027.1 help documents and read-only searches of a detected local installation, with an explicit help-only fallback.