Plugins

Data asset attestation plugin that generates certification-level data quality reports with evidence chains.

Re-speaks each assistant report in a chosen persona outside the working prompt, with four built-in presets, per-session switches, and the original available on demand.

Data visualization plugin that generates self-contained HTML reports with five chart types (bar/comparison/pie/radar/flowchart) and two-column layout.

A long-form Chinese novel workbench inside DeepSeek Harness: read, edit and annotate the manuscript in place; AI reports problems with ready-to-adopt revision drafts, and the text changes only when you accept. A ledger for characters, facts and foreshadowing plus a world map keep a long novel consistent, and AI can continue, expand or rewrite chapters.

Chinese metaphysics (xuanxue) divination panel and agent tool with nineteen systems: eight chart systems (BaZi, Zi Wei Dou Shu, LiuYao, MeiHua, Da Liu Ren, Qi Men Dun Jia, compatibility charts, Tarot) plus eleven quick references (almanac, solar–lunar conversion, date selection, bone-weight fortune, Xiao Liu Ren, horoscope, numerology, life-curve, feng shui, name analysis, an MBTI-style 16-type questionnaire), with ask-driven, report-style interpretation. BaZi adds exact dayun handover dates, annual suiyun hints, "passing-through" (通关), three-book classical cross-reference, four-mausoleum (杂气) month pattern selection and daily "human element" (人元司权); Zi Wei adds 34 patterns, eleven switchable school variants (including an opt-in full 61-star minor-star table that ships the placement verses alongside the chart), chained flying transformations and three-way-four-corner composite flights; LiuYao adds yongshen strength, whole-hexagram per-line strength, hidden-spirit release conditions and response-time clues (应期), while LiuYao and MeiHua quote the received Zhouyi text in full (64 hexagram statements, 384 line statements plus 用九/用六, cross-checked across three editions with variant readings in brackets); feng shui covers eight-mansion and Xuan Kong flying stars; name analysis ships a 1,062-character library with five-grid (五格) numerology and phonetics checks; tarot ships a card-by-card library for all 78 cards (Rider-Waite-Smith base: English name, card-image description, core meaning and astrological correspondence, cross-checked against three published sources — court cards deliberately carry no zodiac because two published mappings conflict, and the chart says so). A classical-source library now backs the readings: 464 indexed citations from 66 classical works, with 584 verifiable quotations including all 120 Qiongtong Baojian pattern extracts; every quote is checked as a literal substring against a 341-page offline source snapshot, variant readings are kept side by side rather than harmonised, modern editorial notes are kept separate from the classical text, and the quotations with their plain-language glosses live only in a standalone citation index page instead of the browser chunks. Every result can carry a reproducible shadow-chart ID (same normalized inputs → same ID; version bumps auto-diff chart fields) plus optional step-by-step deduction traces (rule, inputs, result and source per layer). The browser client ships as a small first-screen entry plus per-tab engine chunks loaded on demand (first-screen gzip ≈ 90KB), while the offline page stays a single file. The DSH panel and the offline page share the same grouped first screen (ask / read a person / pick a date) with recent-used entries and narrow-width folding; charts open with a plain-language conclusion, one-click copy of an ask-ready prompt and saved birth charts, and an unknown birth hour falls back to a three-pillar chart instead of silently assuming midnight. All computed locally; networking is off by default. The npm package also ships an offline single-file page (命理排盘.html): double-click to open, every tab included, no server and no network needed.

Study workbench for exam prep and research: three-column task board with tags and filters, a spaced-repetition queue (simplified SM-2) that keeps finished cards coming back, a heatmap lit by actual focused minutes, weekly reports, per-day review notes, a pomodoro timer that survives panel switches, and Markdown export/import for Obsidian. Installable via dsh plugin add (dsh-study-desk on npm).

Quantitative data quality scoring based on GB/T 36344-2018, providing auditable quality reports for data asset capitalization.

Captures the agent's own self-reported mistakes into a categorised lesson library that is re-injected into the system prompt every turn (each captured draft is confirmed or edited by the user before it is recorded), and adds two conversation-header panels for editing the cross-conversation global prompt and a per-session prompt.

Windows Computer Use for DeepSeek Harness with 21 tools and verified targeting: controls are addressed through their UI Automation identity and patterns, and numbered marks are checked against the live window before input, so a stale target is refused instead of clicked. A known sequence runs in one computer_batch call, one model round trip; incremental observations, compact receipts (normally no more than about 6 KiB extra), bounded handoff waits (HANDOFF_WAIT_TIMEOUT/HANDOFF_HELD_TIMEOUT as continuations), the desktop-block cycle boundary and a deterministic 100 s owner-scoped leave backstop keep context small and prevent a stuck handoff from waiting out the tool deadline. Approval modes settle an unanswerable ask honestly: in a danger-full-access session with no approval card the action runs under that grant, every other sandbox — and every delegated subagent session — refuses and names the unavailable channel instead of reporting a user rejection, and gate diagnostics land in policy.jsonl. Human handoff, a model-visible stop brake and prebuilt native workers.

Evidence-driven goal verification plugin for DeepSeek Harness. Locks goals via /ultragoal and verifies completion against command exit codes and generated artifacts, preventing the model from falsely reporting task completion.

Logged-in browser automation with 170+ site adapters as deterministic one-shot commands (requires the OpenCLI CLI, npm i -g @jackwener/opencli, and its BrowserBridge browser extension) — action-evidence verification (no silent failures), site knowledge & health cards, watch monitoring, trace replay, provenance reports, and a write-approval gate.

Manages OpenCode Zen free-tier and Go-tier models in settings.yaml — fetches the live model listings from opencode.ai, reports configured-versus-live drift per route, and adds or removes entries through four agent tools and an OpenCode Models settings section.

Reference verification for DeepSeek Harness (参考文献真实性检查器): the reference_checker model tool checks a paper's bibliography (.pdf/.bib/.tex/.txt or pasted text) against Crossref, OpenAlex and arXiv, reports found / partial / not_found / error per entry, and emits corrected citations in the input's style (APA, GB/T 7714, IEEE, Vancouver, MLA, Chicago, Harvard, ACS, BibTeX).

Deep-collaboration writing agent for DSH: per-user writing-style learning from edit pairs, clarify→outline→section-writing→red-team→reader-gallery flow, project-to-report synthesis (lab report / product intro / survey / README), 41 MCP tools, and web selection-improve UI with in-app file preview.

A QA plugin for DeepSeek Harness that runs end-to-end checks on a project (build, typecheck, lint, tests, coverage, dependency audit, secret scan) and writes a tickable report with P0–P3 severities and quality gates; execution is limited to preset commands and probing to loopback addresses.

Prompt-driven code audit tasks for DeepSeek Harness with vulnerability tracking, live progress, and exportable reports that include PoC request examples.

Batch pseudonymization for medical text, CSV, JSON, and XLSX in DeepSeek Harness. Deterministic identifier rules run first; an optional OpenAI-compatible model pass adds discovery and verification. Exact-span replacement preserves file structure, tokens stay stable per value, and each batch produces an audit report plus a re-identification vault. DICOM images get header-only de-identification based on PS3.15 Annex E, with pixel data left byte-identical.

Claude-style projects and threads: a project's main conversation opens threads (real sessions you can open and talk to) for self-contained tasks, runs up to 3 at once by default, optionally each in its own git worktree, and gets each thread's report back.

Connect several OpenAI-compatible gateways at once: each one is its own group in the model picker, can be hidden without losing its settings, and can fetch its own public info from its address, alongside per-gateway model management, model tests and optional quota reporting.

Watches a turn for long silences and, once the threshold passes, injects a user message asking the model to state in prose what it is doing before carrying on, so the turn keeps going. The same text is mirrored in a lower-left overlay above the composer. The silence threshold and the minimum gap between reports are set in minutes, the prompt can be replaced wholesale, and it runs with no runtime dependencies. Requires DSH 0.2.1+ and Node.js 18+.

Multi-agent orchestration preset: a lead agent dispatches subagents from a built-in role roster over separate read and write lane pools with per-role model binding, and collects each subagent report on a shared board; adds go_work/continue/need_help/forward tools, a tree-view panel and a settings page.

Reports which PowerShell the DSH pwsh executor actually runs (enabled rows only; version-numbered Store paths are never recommended as pins), and installs a Store-managed PowerShell 7 through a scripted Store-page click or the manual Store path. Allows the pwsh tool only when the live executor resolves PowerShell 7+, and when dsh-ops is installed keeps the native pwsh beside ops_bash by enabling Compatibility and hiding duplicate official tools. Native status card and configuration switch in the Web UI; Windows only, no runtime dependencies.

Skill that types the request before it searches: it splits a question into premises, definitions and analysis units, decides which report skeleton fits (verdict, explainer, causal mechanism, comparison, survey map, metric card), confirms the direction and runs a feasibility check, then emits that skeleton instead of forcing every question into a true-false verdict — in Chinese or English. Evidence gathering never trusts model memory: it searches domestic and international ecosystems in parallel, routes each point to the primary-source ladder for its domain (policy, academic, medical, financial, disaster, food safety, science, education, sports, entertainment, judicial, and more), fetches primary pages, grades sources by type, re-searches with debunking keywords, and keeps every statement bound to a full URL and citation position. Three factual judgements can be routed to a TypeSafe Jev (System One) decision model when a key is configured, falling back to the skill's deterministic rules when it is not; a model verdict can only lower a confidence tier.

Plan-usage pet for the DSH web GUI: a floating skeletal-animated character reports quota in a speech bubble, auto-matching the configured provider (OpenCode Go, Command Code, DeepSeek) to show a plan percentage or an account balance, and falling back to local token stats when the provider API is unavailable; draggable, wheel-scalable (90-260px), synthesized click sounds, loopback-only routes.