Плагины

File-defined command hooks across the session, prompt, tool, permission, Stop and PostCompact seams, with an approval-only GUI: no unapproved command runs, and approval binds the exact definition, canonical working directory and pinned watchFiles. Commands run as the DSH OS user without a hook sandbox, so the trust ledger is explicit.

Source-install updater for DeepSeek Harness: checks the official GitHub release and the npm record, builds the target tag in an isolated directory, smoke-starts it, then switches through a supervisor process with rollback, with scheduled checks and a daily idle-only install window. Source install only (macOS/Linux); requires Node.js 22.19+, pnpm and Git.

Connects DSH to local HTTP proxies automatically via a loopback relay. No TUN, no env vars, zero-restart switching, with sidebar/composer dock controls and live topology.

Safe detached DSH restart: /restart and /dsh-restart slash commands with native candidate styling plus a dsh_restart agent tool; a detached worker restarts DSH after 3 seconds without freezing the session, and Windows/macOS use a native Node helper built on the official appExit service.

Panel entry for the official client-plugin hot reload that tells you what actually changed first: it diffs each row rev against the last seen state and reloads only the plugin bundles that changed, keeps whole-graph reloads behind an explicit confirmation (a bare call is refused and reports the affected row count), and reports replay counters so a silent no-op cannot hide.

DeepSeek Harness plugin for rulemux: asks the local rulemux CLI for the rules rendered for this workspace (rulemux inject --agent dsh) and injects the output into the session at the first user turn, once per session. Nothing is written into the workspace for rules; skills are the only part that lands on disk.

Personal WeChat and WeCom (企业微信) bridge: QR pairing, one DSH conversation per chat, two-way text/image/video/document/archive transfer, chat commands (/help, /status, /new, /stop, /friends) answered by the bridge itself without spending a model turn, a switch that restricts the bot to 1:1 chats (group messages are dropped silently and counted), remote approval by replying 1 or 2 in the chat, and a two-way WeCom intelligent robot over its long connection (no public URL needed) that also receives and sends images, files and videos.

Runs a diff past several reviewers at once — a CLI coding agent such as codex or claude, or a DSH subagent — and merges their separate sections into one report; reviewers and named review modes are configured in the settings UI.

Git workbench in the DSH session header and overlay: view changes, commit, fetch, pull, push, and browse branches and history in existing local repositories.

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.

Runs a second agent: hands independent subtasks to a local Codex CLI in parallel (up to three at a time), and asks it to review a design or a plan before the user sees it, returning conclusion, evidence, counterexamples and a verification step. A dispatch missing its timeout or its acceptance check is rejected by the plugin, not by the model remembering to check.

Conversation search, checkpoints, context management, main and side chats, local long-term memory, tasks, and daily planning for DeepSeek Harness.

Agent preset for DeepSeek Harness: the standard tool stack plus four working rules written into the persona prompt — a project keeps at most four documents (intro, decisions, change log, current state and next steps), those documents describe only the current state with no AI edit history or superseded decisions, no git commit or push or pull request until the user explicitly asks, and a careful-coding workflow (restate the goal and done-criteria first, surgical diffs, evidence before claiming success) with plain-language reports.

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

Deterministic step governance for DSH agents: rejects over-budget agent steps and denies risky tool executions before they run, with shadow mode by default and fail-open on errors.

MCP bridge that lets Codex start and resume DSH headless sessions through three tools: dsh_env, dsh_run and dsh_resume.

Windows-only. Trust a newly added workspace folder once, then grants the NTFS rights the workspace-write sandbox needs and pushes the grant plus the Low integrity label through folders that already exist; repairs show live progress.

Keeps a DeepSeek Harness run going on its own: continues a turn that ended abnormally, and continues a turn logged as completed while the model tool call was emitted as plain text instead, which leaves the agent silently idle. Revives a model request that stops streaming. A wait on an approval or on a question is logged and notified, never answered.

Registers Epic's Unreal Engine skills (unreal-mcp, create-toolset, unreal-skill), injects Unreal project context, and connects the editor's ModelContextProtocol server with a port form on the Plugins page.

Adds a writing-plans skill that generates a structured implementation plan document before coding, then executes simple tasks inline or offers an execution choice for complex work.

Doxygen comment workflow for C/C++ (including Qt) sources: file headers, functions, module groups, trailing and short comments, plus a scan that flags comments whose signature fingerprint went stale. Requires Python 3.10+ with `pip install tree-sitter tree-sitter-cpp`.

Skill panel: browse skills by source, group them into colored folders, preview files, export a skill as zip, copy one into the project, rename it on disk, and insert /skill calls from a composer menu.

Eight academic skills bundled for one-command install — publication-ready academic figures designed from real data with visual review, AI-text humanization, paper polishing and rewriting, PubMed citation verification (DOI/PMID reconciliation), Chinese citation and medical-book reference checks, and WeChat medical article structuring. Three skills run fully local, five call external services only when used (NCBI E-utilities, DOI resolvers, publishing platforms), declared per skill. Zero telemetry, no install-time scripts. Skills run local Python 3.8+; the academic-figures skill also needs its pip dependencies installed.

Kill AI Slop for DSH — score a received document on three weighted axes (expression, execution, decision), audit work-slop with a delivery-obligation ledger, and machine-verify every quote against the source. Reads docx, pptx, xlsx, pdf, txt, md, csv, json and html. Zero runtime dependencies, zero network, fully local.