Skip to main content

Plugins

Browse, filter, and install DeepSeek-Harness plugins.

1191 plugins found

C

dsh-prompt-smith

cyberpzy/dsh-prompt-smith

Adds a sparkle button to the DSH composer's tool row that rewrites the current draft into a structured prompt through one auxiliary call over the profile's own LLM route, writes the result back into the input (Ctrl+Z restores the original), fences the draft against prompt injection, and never touches the session log; a settings page selects provider, model and reasoning effort from the host's own model catalog with a test call that reports unsupported effort levels, edits the built-in system prompt (persisted through the host config editor), and tunes temperature, max tokens and max draft length. Zero dependencies, no build step.

03 days agoUI EnhancementsMIT
A

dsh-macos-menu-bar

amoswzw/dsh-macos-menu-bar

Adds macOS menu bar access to the existing DeepSeek Harness Desktop window, with native approval prompts and optional notch status and approval cards.

02 days agoUI EnhancementsMIT
G

dsh-hermes-link

guiyinrenshi/dsh-hermes-link

Bidirectional bridge between Hermes Agent and DeepSeek Harness, compatible with dsh 0.2.x (subagents facade startActivation/prompt, session format v4). Hermes dispatches focused tasks via JSON-RPC over POST /mcp/collab (one-shot or continuable, nonce-boun

08 days agoMCP & ConnectorsMIT
C

dsh-plugin-console

cheemsup/dsh-plugin-console

DSH sidebar plugin console: inspect plugin information, session sources and tool ownership, preview context content and the impact of disabling a plugin, and customize prompts and tools for the first request.

0yesterdayDev & Plugin Tools
6

dsh-disk-cleanup

6starlong/dsh-disk-cleanup

DSH plugin that routes shell temp dirs and pnpm/bun/uv/npm caches out of the workspace, grants the scratch dir write permission, points the model at the temp location via the system prompt, and cleans up expired temp files.

03 days agoTools & Capabilities
0

dsh-memory-vault

07-dot/dsh-memory-vault

Recalls an Obsidian Markdown vault before each task: a per-agent prompt context is registered on agent creation and re-queried with the current task text, adding a standing profile plus the notes that match, while three tools read, write and audit the vault (inbox backlog, broken links, duplicate candidates, stale goals). Scoring is local n-gram matching with no embedding model, no external service and no runtime dependencies.

06 days agoMemoryMIT
1

dsh-handoff-at-75

1502074527-svg/dsh-handoff-at-75

Hands off a conversation when context usage reaches a configured threshold: writes a handoff document into the project and continues the task in a new session in the same workspace, with the new title carrying a -N suffix, an optional translucent four-color status tint on sidebar session rows, and a configurable model + reasoning effort for the handoff conversation (also selectable in a prompt at handoff time).

04 days agoSessions & Messages
Q

dsh-prompt-expansion

qing-gu-web/dsh-prompt-expansion

A composer button, an /optimize command and an optimize_prompt tool that rewrite a prompt into a numbered specification, leaving variables, code blocks, URLs and numbers unchanged.

06 days agoUI EnhancementsMIT
C

dsh-prompt-polish

chuilian/dsh-prompt-polish

Prompt polishing in the DSH composer: a button rewrites the draft with the profile's own model, then shows the original and the rewrite side by side with a list of every edit before anything is sent. The rewrite keeps the intent, names and numbers the author gave, and refuses to invent a detail it cannot know. Its HTTP routes have no DSH access check, so a web page open in your browser could call them.

04 days agoUI EnhancementsMIT
Y

dsh-prompt-upgrade

yiyunet/dsh-prompt-upgrade

Rewrites the DSH composer draft into a well-structured prompt through the harness LLM service, shows a before/after preview, and only then fills it back, copies it, or undoes the change. The original draft is always preserved, and the settings panel lives inside the 「ANB 工具」 settings shell. Fork of dsh-prompt-enhance. Its HTTP routes have no DSH access check, so a web page open in your browser could call them.

06 days agoMemoryApache-2.0
M

dsh-stock-quote

mengmenghao/dsh-stock-quote

Real-time quotes for A-share, Hong Kong and US listings: the stock_quote and stock_search tools, a watchlist page, and a per-stock page with daily candles and an AI analysis prompt.

017 days agoTools & CapabilitiesMIT
S

dsh-element-picker

sakikotgw/dsh-element-picker

Reference chips: one strip icon in the right sidebar turns elements, links and file paths into native composer chips — pick from DSH itself, same-origin frames, or the sidebar browser's webview guests; the chips serialize into the prompt and point back at

05 days agoUI EnhancementsMIT
G

dsh-workspace (dsh-plugin-airlock)

gabrielbav/dsh-workspace

Puts every plugin, skill and subagent bundle through 6 security scanners (Cisco Skill Scanner, SkillScan, Cisco MCP Scanner, OSV-Scanner, Trivy, Gitleaks) and an optional LLM judge on your own DSH model, in a quarantine with install scripts off, checking for prompt injection, MCP tool poisoning, vulnerable dependencies and leaked secrets; then it installs, asks you, or blocks, and confirms the installed files are the ones it scanned. Works on stock DSH.

07 days agoSecurity & PermissionsMIT
V

dsh-ai-fastvideo

veillessrose/dsh-ai-fastvideo

Text-driven video generator for DSH: writes one Canvas page per scene from 341 bundled style prompts, renders frames with a headless browser, then adds TTS narration, music and sound effects with ffmpeg.

06 days agoTools & CapabilitiesMIT
S

dsh-prompt-library

super-qaq/dsh-prompt-library

Prompt library for the agent and the composer: save prompts, search and filter by tags, insert into the draft or copy, import prompts from a link with the default model, and track skill or plugin dependencies.

06 days agoTools & CapabilitiesMIT
H

dsh-plugin-trim

haidai233/dsh-plugin-trim

Trims the tool surface per session. Each session owns a profile — a tool whitelist — applied through `agent.ctx.tools.restrict({ allow })` on that agent scope only, and switching one session never touches another; `full` is the built-in no-restriction profile. The provider only picks the default: a session created or switched onto a local provider (explicit `localProvider`, `routes`, a loopback/private baseURL, or a local-ish id) gets the `local` profile, a cloud one gets `full`, and the re-plan runs on model-selection events plus one delayed pass because the projection can land after the event. Profiles are editable from the panel and persisted under `~/.dsh/plugin-trim-state.json`; an explicit per-session choice overrides the provider default until cleared. A separate, opt-in layer physically disables whole plugin entries through `pluginManager.setPluginEnabled` — that one is profile-scoped by API and the panel labels it as such rather than pretending it can be per-session. A `tool_trim` tool exposes list/status/profile/set/apply/restore to the model. The panel is served by a loopback-only HTTP server the plugin starts on its own fixed port — `http://127.0.0.1:12256/` (auto-bumps on collision, configurable) — because the DSH port changes on every restart and a locally linked package gets no market card in the sidebar; it lists live sessions with their profile and tool count, an editor per profile, a provider picker limited to providers declared in the profile or currently in use, and the full plugin list with search and paging. Two implementation details worth knowing: `tools.restrict()` is an intersection and cannot widen, so the old restriction is released before the visible set is read — otherwise switching to a bigger profile would shrink instead of grow; and every disposer is registered on `ctx.effect`, without which a hot reload leaves an unremovable restriction behind. Measured on a 12GB laptop GPU with a 27B model and a fully resident 256K cache, same conversation: 105 tools / 56,306 prompt tokens / 22.5 tok/s decode, 62 / 34,207 / 43.0 after disabling three tool-heavy plugins, 14 / 12,890 / 66.8 after a session moved to the local profile, with prefill rising from 487 to 1,310 tok/s. Restricting also drops the skill catalog injection, roughly 4.3K more tokens. Its HTTP routes have no DSH access check, so a web page open in your browser could call them.

07 days agoTools & CapabilitiesMIT
M

dsh-server-memory (plugins)

methpia/dsh-server-memory

Self-hosted project memory for DSH: a compact prompt index, search and save tools, and up to one consolidation per day written back to your own memoryd service, which runs on your laptop, a LAN box or a cloud server and never calls an LLM itself.

05 days agoMemoryMIT
L

dsh-prompt-filter

lihongxu0221/dsh-prompt-filter

Rewrites DeepSeek Harness identity text and tool names out of requests sent to selected models, restoring original names in responses so tool calls still run.

05 days agoIntegrations & RemoteMIT
S

dsh-opencode-go-session

sw503544488/dsh-opencode-go-session

Sends a stable per-conversation x-opencode-session header on every pi-ai call to OpenCode through the harness header hook, so routing and prompt caching stay per conversation. Its HTTP routes have no DSH access check, so a web page open in your browser could call them.

06 days agoModels & ProvidersMIT
P

dsh-context-ruler

pcysanji/dsh-context-ruler

Shows the live payload size of the current conversation on the answer action row, with warn and danger colors and a one-click handoff that writes a handoff document and leaves a prompt in a new session. Its HTTP routes have no DSH access check, so a web page open in your browser could call them.

06 days agoUsage & BillingMIT
A

dsh-loci

awenforever/dsh-loci

Layered long-term memory for DeepSeek Harness: the prompt sees only a stable index, full text is one lookup away. Bi-temporal facts, strict scope isolation, verified-only injection.

06 days agoMemoryMIT
G

dsh-auto-review-model-select

gt-610/dsh-auto-review-model-select

Gives Auto review its own model instead of inheriting the main one, and repairs replies that break the reviewer's closed JSON protocol: prose, Markdown fences, an `allow` carrying a `reason`, a truncated finish, or a decision emitted as a tool call no longer become `AUTO_REVIEW_FAILED`. Contradictions resolve fail-closed (`high + allow` becomes a denial), and an unreadable reply degrades to an ordinary approval prompt instead of blocking the call.

06 days agoSecurity & PermissionsMIT
J

dsh-read-it-later

jes-phd/dsh-read-it-later

Send the prompt, reply and attachments of each conversation turn to your own Feishu or WeChat bot, for later reading, auto-send after a queued prompt, and cross-machine handoff.

06 days agoIntegrations & RemoteMIT
L

dsh-auto-wake

lzlrt/dsh-auto-wake

Attaches a wake rule to any session: the host polls whether its latest turn ended cleanly (LLM error, max-tokens truncation, unclosed turn, prolonged silence) and genuinely wakes it through sessionController.prompt, with a Beijing-time window, an idle-delay gate and three guards (cooldown, hourly/daily quota, no-progress circuit breaker). Includes five agent tools and a settings page. Its HTTP routes go through the host connection trust check.

06 days agoWorkflow & AutomationMIT