Plugins

TypeSafe Jev (System One decision model) for DeepSeek Harness, end to end: jev_decision (choice/score/noul, parallel, typed answers + confidence), a Settings > Plugins card for the API key and switches, inline tool views showing every decision in the conversation (answer, confidence, latency, cost), an auto-guard (deterministic + Jev risk/loop checks), a local /jev dashboard, and an online verification benchmark (96.3% on 27 labeled questions, median ~283 ms on jev-latest, 2026-09-21). Honest by design: real API only, no mock fallback.

Multi-instance ComfyUI orchestration: probes every configured ComfyUI endpoint for its GPU, free VRAM and queue depth, then dispatches a generation job to the idlest eligible instance. Ships three tools (comfyui_farm_status, comfyui_farm_pick, comfyui_farm_run). An instance is eligible only if it is reachable, has at least the requested free VRAM, and its queue is no deeper than the configured ceiling; eligible instances are ranked by free VRAM minus queue depth times a weight, so an idle shallow queue wins over a busy large one. Zero runtime dependencies. Optional retry on probe failure, because a single timeout across an SSH tunnel is not proof that an instance is down.

Content-degradation detection (thinking loops, output repetition) for DeepSeek Harness, with a continue-vs-handoff cost verdict and zero-model handoff documents.

Workspace checkpoints per tool call: records every file mutation inside the workspace (excluding .git and node_modules by default), including changes made by shell commands rather than by write/edit, and stores the previous content of each changed path in a content-addressed store kept outside the workspace. Restores have two modes — undo one record, or return to a point in time — with a line-level diff preview, a dry run by default, and a rescue record written before every restore so the restore itself can be undone.

RainbowSpeak — a flowing rainbow marquee border around the chat composer card, driven by a conic-gradient ring. Live settings for speed, width, seven color schemes, direction, sides and breathing, plus session-state awareness that reacts to running, attention, error and celebration moments.

Searches wallhaven.cc from a settings page and sets any wallpaper as the DSH web GUI background, with opacity, blur and scrim controls, a sidebar shuffle button and one-click original download; all fetching runs in the host process, so the browser never needs to reach wallhaven.

Skill that fact-checks a claim against the open web rather than model memory: it splits the claim into independently checkable fact points, routes each point to the primary-source ladder for its domain (policy, academic, medical, financial, disaster and accidents, food safety, science, education, sports, entertainment, judicial, and more), searches domestic and international ecosystems in parallel, fetches primary pages, grades sources by type, re-searches with debunking keywords, and reports conclusion-first with a full URL and citation position behind every statement, keeping unverified and undecidable points separate from confirmed ones. 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.

Task monitor plus remote chat for DeepSeek Harness. Read sessions, background jobs, subagents, goal rounds, workflows and pending approvals from a phone-friendly page, and talk to any session from it: streamed replies, tool-call progress, stop at any time, switch sessions and models. Also ships a GUI sidebar panel.

Blocks an agent from reading, searching, writing, or editing chosen files and directories, keeping their contents out of the conversation — including sensitive-word files whose text can make later provider requests fail with a 400. Rules are picked from a file/directory browser on the plugin page.

Filesystem permissions for shell calls, granted per path and capability rather than per command name. The command line is parsed for the read, write, create, delete and execute effects it needs, a capability it lacks raises the approval card instead of a flat refusal, and the same rules are compiled into the macOS Seatbelt profile the process, its children and the code its arguments never showed all run under. An Approvals tab in the conversation reads the audit log back and shows which rule, automatic reviewer or person answered each call.

Fronts stdio MCP servers with three router tools (list_capabilities, describe_tool, call_tool) so schemas load on demand instead of riding every request; shapes every tool result through a deterministic output ladder that archives the original first (retrieve_spill reads it back); prices tool groups in tokens and can hide built-in groups per agent. Bridges to a standard-library Python router, so Python 3.9+ is required.

Mounts the mcptoon MCP gateway as one stdio server, so the agent discovers the tools of every MCP server you have configured through mcptoon's compressed view instead of full JSON schemas on each turn. Ships a doctor CLI that checks the mcptoon executable resolves and prints the exact command/args override when it does not.

Panel to edit dsh's native AGENTS.md from the web settings page: every layer in the instruction chain is clickable, three activation scopes, and a visual budget meter.

A dsh-better-sidebar tab that statically parses Qoder Canvas `*.canvas.tsx` files into structured pages in the DSH right sidebar (the source is never executed: no eval, no bundler, no sandbox iframe), with a file-viewer takeover for `.tsx` and a bundled Skill for authoring the format.

Simulated life context for DeepSeek Harness: every turn injects the past 24h of simulated life events from the workspace .life/ directory (sliding-window filter, session-level deduplication with delta updates, world summary) as a collapsed context notice, and the life_react tool lets the agent record feelings, thoughts and actions back into its life log for the next round of world evolution — turning a DSH agent into a roleplay character with a persistent daily life.

Compaction fallback for compaction-basic: when its LLM summary throws, comes back empty or cannot fit the window, the span is compiled deterministically with a seq pointer on every line (vendored dsh-compaction-instant compiler). Adds recall and search tools, /recall and /context commands, an idle trigger at 45% of the window and an on-disk span archive. The engine mounts as one row in the agent preset (docs/dsh-integration.md).

Humanized fork of @xmanrui/dsh-im (base 4.21.2): humanized message delivery with send-delay and typing two-phase, message_break + streaming coexistence, status reactions, reply quotes, per-bot humanize overrides, and a no_reply reclaim tool — across WeChat, Feishu, DingTalk, WeCom, QQ, Telegram, Slack and WhatsApp channels.

In-page task notifications with collapsible status toasts, sound, and session navigation.

A DeepSeek Harness bundle of the 25 promoted mattpocock/skills with a native agent preset, an invocation-aware skill provider, and a light grilling-to-implement bootstrap.

Takes over the built-in web_search and web_fetch providers with Tavily, each behind its own toggle: a key pool with balance-aware scheduling that reads real credit balances from /usage, failover with hard exclusion of cooling or exhausted keys, and call history with a 14-day credit chart.

Bridges Feishu, WeCom, DingTalk, QQ, Slack, Discord, Telegram and Mattermost into dsh agents (outbound WS/Stream/Gateway/long-poll/webhook), with im_status, optional local Whisper ASR, plain-text outbound for QQ/DingTalk/Telegram/Mattermost, and env CSV allowlists (DSH_IM_*_ALLOWED_USER_IDS). Empty allowedUserIds means EVERYONE can drive your agent — set a whitelist before exposing a bot.

Model-autonomous context compaction for DeepSeek Harness: a mind_map tool surveys the agent context surface (stable seq ids, roles, token weights, clear-boundary markers, the clear-mind playbook), a clear_mind tool replaces a chosen conversation-history span with a model-written checkpoint through the native compaction transaction, call and result pairs self-collapse into one-line tombstones, and a proactive reminder nudges the model when the context grows long or a turn runs too many steps.

Model-callable compact_agents tool that force-compacts context in every live session (main session, ordinary sub-agents and AgentTeams members alike), queues busy targets and compacts them when the turn ends, reports the shadowed node count and estimated tokens per target, and ships a browser form for the trigger ratio, retained ratio, controlled-phase output budget, auto-continue budget and an end-of-turn pre-compaction ratio that compacts near the trigger line once a turn ends, so a reply no longer opens with a wait for the summarizer.

One settings page that lists every subscription sign-in the harness can offer and walks you through it — ChatGPT (Codex), Claude, GitHub Copilot, OpenRouter, Kimi and xAI — reading the list from the authorization seam instead of hardcoding providers, and showing credential records no registered flow claims any more.