Plugins

Skill Manager page in the DSH settings panel: browse system / user / workspace / preset skills, expand a skill to its file tree, view and edit files, import skills from a zip, and export or delete them (system skills read-only).

Vision toolkit for text-only DeepSeek: model-invokable `vision` tool, wrapper adapters for deepseek/opencode-go (v4 flash/pro), Antigravity IDE quota (default, flash/pro) / any OpenAI-compatible VLM / Gemini / local Ollama channels, evidence memory with compaction rehydration, content-hash cache, and a bilingual client panel.

Real-time DeepSeek API usage: account balance with low-balance alerts (browser notification on threshold crossing), token & request stats (total/today/last-60s windows, per-model and per-session), estimated cost — a composer dock plus a settings-page usage panel.

MiniMax token-plan quota dock for the Web UI: a drop-shape 5h-usage % chip in the conversation input area, click for a detail panel listing every model's 5h & weekly usage and reset countdown, auto-hidden when the active default model is not a MiniMax model.

Download progress panel: URL downloader, agent shell/SSH transfer tracking and workspace black-box file-growth monitoring, shown in a draggable floating panel with live bytes, speed, percentage and ETA.

Skill manager for the dsh web GUI: a sidebar entry plus a panel listing the skills visible to the current session (split into global / local groups) with per-skill on/off switches, create / import / edit / delete. Disabling a skill removes it from the mod

Right-side developer panel for DeepSeek Harness: files, session history, and git version control

Collapsible iOS-style stats pill under the composer: session cost, DeepSeek account balance, cache-hit rate, and token usage in a frosted panel.

Control labscript-suite experiments from DeepSeek Harness: runmanager globals and scans, submits and waits, YAML fixed-operation protocols with postprocess write-back, panel snapshots, a built-in mock backend and a read-only doctor.

MCP panel for DSH: turn a whole server on or off, hide individual tools, and see how many context tokens each server's tool definitions use. Its HTTP routes have no DSH access check, so a web page open in your browser could call them.

Code can Vibe, progress must be Visible. No guess, less stress. See progress you can assess. Make project progress visible with local algorithms, optional AI assistance, task checklists, and subagent tracking. View overall progress, subagent activity, current operations, and estimated time remaining in a draggable panel. Supports DeepSeek Harness Web and Desktop. Requires Harness ≥0.2.2-rc.1 and Node.js ≥24.9.0. Local estimates require no additional model requests. AI assistance is off by default; enabling it incurs charges from the selected model provider. Probe usage and estimated costs are available in the panel.

Build named subagents and skills in a sidebar panel, then call any subagent by name in a conversation. Each subagent selects a model, reasoning effort, spawn or fork mode, a tool allowlist and skills; skills can be global, per workspace or per conversation.

Inspect API latency, streaming output and model-response fingerprints from a native DSH panel, with classic-task previews and public benchmark rankings.

Configurable context compaction with three independently switchable sections — whole-context checkpoint, reasoning compression and tool-output pruning — each with its own pressure threshold and method. Mechanical modes make no model calls; model modes summarize fragments in batches with a configurable provider and prompt. A left-sidebar panel exposes switches, method cards and editable parameters that save without a restart, and every rewrite is archived with its original text.

Browse local AI agent projects in one DeepSeek Harness panel: filter and pin projects, pick memories, skills, and recent sessions, then continue in the original directory with an editable DSH handoff draft or export it as Markdown.

Floating HUD for the DeepSeek Harness Web GUI: live session token speed, input/output/total token counts and cache hit rate, DeepSeek account balance, and per-model cost estimates in a draggable, resizable panel.

Drag any sidebar panel to reorder it, or drop one onto More to fold it into the More submenu; the order and the folded set are remembered across restarts.

Pre-execution review gate for file writes, edits and deletions: the tool call is held before it runs, a right-sidebar panel shows the computed diff (unified, or source versus result side by side), and nothing is written until Accept is clicked; pwsh commands that create, overwrite or delete files through a literal path are held the same way, and a command whose target is only known at runtime shows its raw command line instead of a diff. A command touching several files is held once and reviewed as one tab per file. What a diff cannot express is reported instead — a binary file or source map shows its byte sizes, and a change that only touched line endings or indentation says so. Accepted changes are kept as snapshots and can be restored from a history view, and tools without a dedicated adapter are judged by their arguments: a path plus a mutating verb is held, a read is not.

Runtime monitor for the plugins installed in a DSH profile: the panel lists every bundle and plugin entry with its Cordis fiber state and load diagnostics, scores risk from failed entries, missing bundles, version exemptions, install scripts, bin entries, a missing license and non-registry sources, keeps a timeline of state changes, and can disable a plugin entry or a whole bundle without leaving the panel.

Lightweight, self-contained desktop pet for DeepSeek Harness — an in-app floating pet driven by session activity, a settings panel, and a transparent always-on-top window that reuses the DSH host's Electron runtime (path resolution mirrors dsh-pet).

AlertManager panel in the DSH sidebar: reads multiple AlertManager HTTP API v2 instances in parallel, aggregates their alerts into severity-colored cards filterable by search, state, severity and instance, and dispatches any selection into a new session in a chosen workspace; the instance-grouped alerts become the first message to the agent, which then starts handling them immediately. Supports per-instance auth and TLS, auto-refresh, and a configurable dispatch prompt.

Read Miniflux inside DeepSeek Harness in a three-column panel (feeds, article list, article) and ask the assistant for unread digests, per-category briefings, full-text reads and read/star actions on the same account. Needs your own Miniflux server (2.0 or later; 2.3.2 for batch starring) and an API token, kept in the Harness credential vault (the MINIFLUX_BASE_URL and MINIFLUX_API_TOKEN environment variables also work).

Derives the index from your Markdown memory files on every turn instead of keeping a hand-written index table, so renames, splits and deletions never leave a stale entry. Retrieval ranks by library tags, section tags, title tokens and body text, then injects the matching line ranges rather than whole files. Writing is model-proposed and user-confirmed through a sticky note; entries are appended after the target section's format is detected. Ships a structure panel showing each library and section with line ranges and token estimates, a starter template for an empty memory directory, and optional appearance controls.

Integrated control plane for the local strata model stack: manage the strata-hop :8080 model-switching router (status/models/warm/logs), stream live model-switch and load state into the web GUI status chip and the strata_status tool, preload a model engine the moment it is selected, edit run-config and shared chat settings from a GUI tuning panel or strata_params, and ship the dsh provider patch that routes chats to the router with Effort-level mappings. Read-only toward inference: it never sends a prompt and spends no tokens; bring your own running strata plus strata-hop (defaults :8080 hop, :8756 monitor). As shipped, the status server listens on 0.0.0.0:8756 with no authentication and open CORS, and its endpoints can change run-config and restart the engine; set host to 127.0.0.1 in cordis.patch.yml to keep it on this machine. It expects strata-hop as a systemd user unit.