Plugins

Adversarial review gate for tool calls. It hooks `tools/pre-execute`: a local regex pass over destructive commands (`rm -rf`, `git push --force`, `git reset --hard`, `DROP TABLE`, `format`, disk writes) and sensitive paths (`.env`, `id_rsa`, `.ssh`, `.git/`, `$DSH_HOME`) runs first with no latency, and only a hit is escalated to one review-model call that returns allow or deny. Plain reads and ordinary writes never enter that path, so everyday work pays no added delay. Every verdict is shown live in the Web GUI.

Writes a model-generated annotation into every DSH escalation approval prompt — what the command does, what the model intends, how risky it is and why, and what each choice means — and keeps a searchable history of past approvals at /dsh-approval/history, with a diagnostics page at /dsh-approval/diag.

GitHub issue service for DSH: turns a telemetry failure analysis into a uniform issue report, builds a pre-filled new-issue URL that is shortened when it would exceed a configured length budget, and rewrites a plain description into a structured issue with the deployment's model.

Routes eligible bash commands through rtk (Rust Token Killer) so tool output is compressed before it reaches the model, by asking the installed rtk to rewrite each command and adopting what it prints; when rtk is absent the executor behaves exactly like the stock one.

Records every session as one readable Markdown timeline inside the session workspace - the full request sent to the model (system prompt and plugin injections included), user input, replies, reasoning, tool calls and results, file reads and writes, web searches, skill loads, command runs and token usage - with a review page, a per-session tab and configurable capture scope.

Agent Studio for DSH Web: configure models, prompts, per-plugin tool bindings and Skills; trial-run with editable tool results and run history; preserve session capability snapshots; export standalone presets; draft Agents with a guided creator preset.

Jenkins CI integration with a live build-progress panel in the right sidebar, a multi-instance settings page, five model-facing tools (list jobs, build status, console log, workspace, trigger build), and a failure wake that hands a failed followed build to the conversation that followed it.

Zagent — unified three-role self-improvement loop with three operators. Data-RSI (daily cross-sectional cleaning: winsorize → cross-sectional standardization → median fill, plus experience extraction), Harness-RSI (evolves system prompt / memory / skills), and Model-RSI (evolves alpha training recipes via framework-native walk-forward with 2-day gap and multi-seed). Offline one-command run, GPU auto-detect. Ships one skill.

Host-side long-task detection for small-context local models: long outputs are scaffolded into a parts skeleton (one part per reply, then assemble), and multi-step work into a PLAN.md ledger with notes/ (plan first, one step per reply). Detection is heuristic and host-side, so it does not depend on the model judging its own limits.

Persistent Bots and group rooms inside DSH — each Bot keeps its own model, permission preset, workspace and session across restarts, and every answer is a run written to a ledger the sidebar panel replays.

A resident assistant for DeepSeek Harness that is deliberately not another chat box: one floating window plus one standing conversation, both independent of every other session. The heartbeat is held back by four gates — an authority grant, a daily turn budget, quiet hours and a pacing cooldown — and it keeps working across restarts through a lease it has to hold. Each wake is decided by a model that must answer SKIP or a short message, so silence is a decision rather than a failure. The window shows the standing conversation's real message flow, read from the host session store, or from the on-disk session log when the host does not publish that session, and it can send into that conversation directly. den carries the same reach as the person it stands in for, with one limit: read-only. It may read every session, this machine and work channels such as GitHub, but it does not write, send or publish, it never asks anyone for permission, and it never gates another session's tool calls. `npm run digest:github` turns GitHub notifications and your own open pull requests into three Chinese sections — waiting on you, waiting on your review, and yours — with reply drafts that are explicitly marked as unsent.

Daily and weekly task manager: startup popup reminder, a marquee strip above the composer, one-click circle completion, editable task cards, and a Sunday AI weekly summary, plus a sidebar DeepSeek API usage widget with balance card, today/cumulative tokens and cost, per-model breakdown and a 7-day chart. Persists as JSON under $DSH_HOME.

Answers a hard question with a one-page visual HTML explainer: the model writes a short Markdown draft and the bundled dependency-free `am` CLI builds the layout, CSS and diagram coordinates. Vendors the upstream answer-me-with-html skill verbatim as a skill-only DSH bundle with no Host code.

Adds an Omni-modal settings section to DeepSeek Harness that points the harness at any OpenAI-compatible image and video endpoint: switchable config profiles with separate endpoint, API key, model, path and size for each side, free-form extra request parameters, a connectivity test that does not spend generation quota, and three model-visible tools — generate_image, generate_video and check_video — with async video tasks that survive restarts.

Removes the model-capability gate that blocks image prompts: patches llm.resolveModelInfo results and adapter model tables to declare image input, so routes with no declared input modality can send and read images. No app.asar or config-tree edits.

Pixel-level screenshot diff for visual regression. Two local PNGs in; a machine-readable first line (`PIXEL_CHANGED=true|false`, greppable as an assertion), then the difference ratio, per-region bounding boxes, maximum channel delta and mean absolute error, plus a self-contained HTML comparison page. An ignore-region list with absolute or fractional coordinates excludes clocks, cursors and other per-render noise. It deliberately does no content recognition: it answers only "did this change", which is the part that can be made deterministic. Entirely local — no model call, no network, no upload.

Adds a right-column panel that compares a set of images the user picks from the workspace — 1/2/3/4/6 per page, per-image zoom and pan, the zoom and viewed position of one image synced onto the whole set, and a reorderable filmstrip. The browser loads every image, so none of it reaches the model.

OpenAI-compatible web search provider for the built-in web_search tool: one baseURL/apiKey/model config powers Qwen Token Plan, OpenAI and any Responses API gateway.

Bridge between DSH and the local WorkBuddy desktop app, in four parts: read, search and append WorkBuddy memory, skills, plans and daily logs; run tasks through the local CodeBuddy CLI with four permission profiles; drive the running WorkBuddy window over CDP (open a conversation, read its replies as sanitized rich HTML, type and send, switch the account-billed model, relaunch it with a debug port when it is missing); and a sidebar panel that puts that conversation surface inside DSH. 25 tools and 15 HTTP routes in one plugin.

Translate between Chinese and English without leaving DSH, using your configured models with no extra API key: streaming output, a customizable prompt, and local history to revisit and re-translate.

Code retrieval engine exposed as an MCP server: four tools (code_search, code_outline, code_index, index_status) over a SQLite FTS5 index with BM25 ranking, CJK bigram indexing, declaration-aware tree-sitter chunking, incremental reindexing, and a three-layer secret filter. The bigram index matches Chinese-language queries against English code, and the default path needs no vector store or embedding model. Requires a local Python 3.10-3.12 engine installed by scripts/install.sh, and CODERAG_ROOT set in the profile patch.

Turn-level file undo for tool writes and edits: each file is snapshotted before the write lands, and the model (an undo tool), the human (a /undo command) or a button in the assistant message action row can roll back every file a turn changed. Snapshots are written in-process rather than by shelling out, so it needs no extra sandbox permission in sessions confined to workspace-write.

Mounts the tsm-core-dlr semantic service (LanceDB vectors + graph) as MCP tools and ships the scenario-independent DLR/TSM modeling skill, a service-status overlay and the agent rules; the service is a hard dependency and must be running before the bundle activates.