Plugins

Per-call model, provider, persona, and toolFilter overrides for subagent delegation, with @preset: references and provider/model composite ids.

Composer-attached images are transcribed to text by an OpenAI-compatible vision model before reaching text-only DeepSeek models.

Bounded local memory with CBDC authority gating: SQLite + FTS5 claims, scoped recall with explainable use/verify/ignore decisions and a full audit trail, /memory commands, ≤3-claim/1200-char injection per call, no extra model call.

A DeepSeek Harness (DSH) sidebar plugin for previewing 3D models, supporting FBX, OBJ, glTF/GLB, and STL files · DSH 3D 查看器

Runs shell commands on machines you have paired with eXe, over its relay, and reads images back from them as image blocks for the model: three model tools (device list, shell, image — take a screenshot with the shell tool, then read the file), plus a sidebar page that shows the caller address of this machine, edits the device and trusted-caller lists with labels, and times a real call to each device on demand. Needs the eXe client running on those machines and a caller key from `exe mcp init` on this machine.

An IDE inside DeepSeek Harness with a file tree, multi-tab editor, AI assistant with model picker and API-key setup, Git panel, PTY terminal, and Python/TypeScript LSP.

Project workspaces for multi-session collaboration, conversation handoffs, searchable chat records, and editable shared memory; automatic memory curation and handoff judgment require two background models (memoryModel and sessionManagerModel) configured in the plugin config.

Second-model gate on the assistant turn, installed in two parts: copy engine/ into ~/.dsh/plugins/ and mount it by absolute path in the profile patch (the bundle itself only installs the settings panel), then switch off the default dry-run mode that only logs verdicts. Once active, a separate judge call scores the draft against the numbered criteria (0-10 each); any criterion under 6 fails the turn and sends it back for a rewrite, up to 3 rounds by default. The web panel edits criteria and thresholds, and every verdict is appended to an audit log.

Registers 12 CatPaw (妙手) models as a DeepSeek Harness provider, with per-model reasoning effort, harness-executed tool calls, and QR-code login. The bundled catpaw-cli is a macOS arm64 binary (other platforms must supply their own copy), and calls use the signed-in CatPaw account quota.

Follow list above the sidebar conversation list: a status dot per conversation (unread, waiting for you, running, failed), newest reply first, running ones last, and a follow toggle beside the model selector.

Redacts secrets, cookies and personal data in every model request at the host process undici dispatcher and restores the placeholders in the reply, so one interception covers every provider, adapters that plugins add included; per-kind actions mask / warn / block, an optional same-shape fake key, records written to disk by the day and a statistics card.

Adds a { } button under each assistant reply in the DSH Web GUI that opens the exact request sent to the model for that reply: model, call parameters, messages, tools, and provider-reported token usage.

Adds right-sidebar document previews for STL, OBJ, PLY, glTF and GLB models, zip/tar/gzip archives, audio and video, fonts, subtitles, TOML, Intel HEX, JCAMP-DX spectra, VCD waveforms, KiCad schematics and boards, and DFU firmware images.

Indexes each chat of a workspace into a small topic graph every 4 turns (one model call, verbatim source kept under .context-graph/), appends up to 5 matching topics from the other chats to each message, searching a named day such as "yesterday" first, and adds a recall tool that returns the original text by topic, chat, day or turn. On and shared by default; /context-graph off and unshare opt a chat out, and stats, build, map and find inspect it.

Adds an /if slash command to DSH: the first line is a condition and the remaining lines are an instruction. A fork subagent that inherits the session tools, system prompt and message history judges the condition and answers YES or NO, and only on YES is the instruction sent to the main agent as an ordinary user message, while a NO sends nothing. The condition, judge prompt and verdict never enter the main transcript. While the agent is busy the command queues with ordinary messages, editable, removable and steerable from the queue panel, and a setting switches to the 0.2.x background mode. The judge runs read-only, and falls back to a one-shot model call over a plain transcript when the host has no context-inheriting fork provider. Supports DSH 0.2.1-rc.5 and later (verified on 0.2.1-rc.5 and 0.2.2-rc.1) and Node.js 20.3 or later.

Prompt booster for the composer. Left-click the composer button and the draft is lightly polished and replaced in place (undoable, no popup). Right-click opens a menu with 8 built-in modes plus your own, where each mode's template can be edited or new modes added (saved on the host), and the model used for the rewrite is picked from the live dsh model list. Other modes show a before/after preview. Based on rongxingda/dsh-prompt-enhance (Apache-2.0); do not install both — both add a composer button and share Ctrl+Alt+E, so disable one of them if the upstream is present.

A session auto-compaction plugin that is friendly to local models: compaction never makes the local engine re-read history, saving a lot of time; the trigger point and the saved content of the compacted session are user-configurable. After installing, point the agent preset's compaction-basic slot at dsh-auto-compaction-engine (or save the settings page to switch it); until then compaction stays native.

Route table assigning provider, model, and reasoning effort to subagents by workspace directory prefix; supports global default policy and in-composer session popover.

Drives the OpenCode client already installed on your machine through its loopback opencode serve API, so its free-tier models work in DSH — the child process starts on first use, and OpenCode executes no tools of its own: a call it proposes is handed to the DSH tool of the same name and run there, or refused when there is no equivalent. No API key is needed.

WorkBuddy models, credits and tool calling for DeepSeek Harness · DSH WorkBuddy 接入插件

Brings the rtk token-saving setup from Claude Code to DeepSeek Harness: commands rtk can rewrite are rewritten, and builds, tests and logs it cannot reach are compressed; git, cargo, npm and pytest are covered. Compressed output tells the model what was elided and how to recover it; when the sandbox cannot reach the rtk store the original is saved and its path given. Savings show without a command: a pill below the composer, with `/rtk stats` for detail. When a rewrite is unreliable the original command runs as usual: if rtk cannot run, or the returned command would drop arguments, nothing breaks. The rtk binary ships with the package, so mainstream platforms need no separate rtk install, unlike most other rtk plugins.

Registers one tool, report_health, that answers whether a scheduled report job actually ran. It judges by side effects rather than exit codes: how old the newest report folder is, how many documents it holds, whether more than one report category is present, whether the audit ledger was synced recently, and how long ago the last log was written. Exits nothing, calls no model, and reads only local paths you can override by config or environment variable.

DSH bundle of the 27 promoted skills from mattpocock/skills v1.3.2 (TDD, code review, grilling, spec and ticket flows, domain modelling), bodies byte-identical to upstream and upstream invocation policies preserved; ships a sync script and an offline contract check.

Compiles a repository into a source-grounded code map: structured pages where every claim carries a verifiable path:line citation, an interactive dependency graph, knowledge cards, four model tools and the /codemap command. Deterministic, no model calls, zero dependencies.