vvlife/awesome-deepseek-harness-plugins
A curated list of plugins, tools, skins, and extensions for DeepSeek Harness (DSH).
安装Install
dsh plugin --profile web add github:vvlife/awesome-deepseek-harness-plugins把 vvlife/awesome-deepseek-harness-plugins 加入你的 DSH 配置（web profile）即可启用。
READMEREADME
Awesome DeepSeek Harness Plugins
中文导读：这是 DeepSeek Harness（DSH）生态的插件精选列表，英文为主维护，中文说明见各章节标题下方。想直接上手的高星插件评测请看 Hands-on Notes。
A curated list of plugins, tools, skins, bridges, and extensions for DeepSeek Harness (DSH) — the open-source agent framework from DeepSeek, built on the motto "Everything is a Plugin."
DSH launched its developer preview on 2026-08-13 (MIT license, Cordis-based). Within a day the community shipped a wave of plugins; this list tracks the notable ones and points to the rest.
Star counts are a launch-day snapshot (2026-08-13) and drift fast. For the unmoderated, auto-refreshed index of every repo tagged
dsh-plugin, see PLUGINS.md (regenerated daily by update.yml).
生态入口 / Ecosystem：想找插件不想翻列表？去 WhaleHub —— 基于本列表每日同步的可视化插件市场（搜索 / 分类 / 一键复制安装命令，还能装进 DSH Web 里点一下直装）。想零配置上手 DSH？试 DeepSeek Harness Desktop —— 自包含 macOS APP（内置 Node + dsh + Paseo），拖进「应用程序」即用。
Contents
- How to install a plugin
- Official built-in plugins
- Community plugins
- Hands-on Notes
- Other awesome lists (meta)
- Contributing
- License
How to install a plugin
中文：DSH 把插件当作 Cordis bundle 加载，最常用的两条路：npm 包用 dsh plugin add <npm-package>，仓库托管（.dsh-plugin 形态）用 github:<owner>/<repo> 形式。
DSH loads plugins as Cordis bundles. Two common paths:
# npm-scoped plugin (recommended)
dsh plugin add <npm-package>
# repo-hosted plugin (the .dsh-plugin format)
# add to your profile's cordis.yml, or via the CLI patch layer:
# github:<owner>/<repo>#<ref>&path:/.dsh-plugin
Start the Web UI and manage models/workspaces there:
dsh web # http://127.0.0.1:3080
Official built-in plugins
中文：框架本体在 @deepseek-ai/dsh-* 这个 npm scope 下自带约 50 个内部插件包，是所有社区插件的参考实现与"接缝"底座（llm / shell / fs / web / subagent / plan / sandbox / hooks / skill …）。
The framework itself ships ~50 internal plugin packages under the
@deepseek-ai/dsh-* npm scope. They are the reference implementations and the
building blocks every community plugin extends. Highlights:
deepseek-ai/deepseek-harness— the framework and all built-in packages. Seepackages/READMEfor the full map:llm(model adapters),shell/terminal/code-runtime(execution),fs/lsp(files & language servers),web(search/fetch),subagent(delegation),plan,sandbox,hooks,skill,compaction,extensions(runtime self-modifying plugins), and theweb/cliapps.
Everything below is community-built and sits on top of these seams.
Community plugins
Web UI & Skins
中文：给 DSH 网页界面换肤、加任务看板、宠物、移动端远程等"界面增强"类插件。
- zhu1090093659/dsh-web-ui (★300) — Plugin & skin collection for the DSH Web UI: task board, git graph, right-side panel, remote mobile UI, pet, live token stats, skin center.
- bpc-oss/dsh-web-billing (★9) — RMB/USD token billing for the DSH web: official-policy auto pricing (incl. peak/off-peak hours), per-message cost ledger, account balance, local-model savings tracking (¥/$ follows the UI language).
- Small-tailqwq/dsh-deep-whale (★56) — "Whale-girl" skin series (maid-atelier), CC BY-NC-SA 4.0.
- Tommy00748/dsh-theme-cyberpunk2077 (★12) — Cyberpunk 2077 / Night City theme: NC yellow × neon cyan, CRT scanlines, Kiroshi hover lock-on, combat-state HUD, synthesized typewriter & message SFX, hidden easter eggs (relic / johnny).
- Nagi-ovo/dsh-ads (★61) — Tongue-in-cheek 2005-style Chinese-site ads in the sidebar / chat feed / popups.
- alingalingling/ui-status-label (★18) — Customize the "deep diving" thinking-status label however you like.
- omdsh-dev/dsh-genui (★9) — GenUI: interactive components (layout, charts, mermaid, 3D) rendered inline via the
dsh-uifence. - vlln/whale-girl (★10) — Desktop-pet plugin (QQ-pet style): draggable, feedable, accumulative companion.
- Nagi-ovo/dsh-visualize (★15) — Generative UI: the model draws interactive HTML cards straight into the chat stream.
- ZSeven-W/dsh-openpencil (★19) — OpenPencil design preview & editing plugin.
- omdsh-dev/dsh-annotation (★9) — Select text → annotate → send as a message; bubble-hidden annotation blocks.
- Anionex/dsh-computer-use (★6) — Computer-use plugin for DSH.
Terminal & Desktop
中文：把 DSH 从网页端带到终端、桌面，或做成独立 App / 启动器。
ccch1mneyyy/dsh-cc-tui (★96) — Claude Code-style full-screen TUI: pixel-whale top bar, live status row, streaming thoughts, double-Esc rollback, context bar + TPS meter. One-line npm install.
huiliyi37/dsh-tianshu-tui (★53) — DSH terminal UI.
chen-001/dsh-grok-tui (★5) — Grok-style TUI.
hust-open-atom-club/oh-dsh-desktop (★46) — Extensible macOS workbench: native PTY, workspace tools, live bilingual plugins, isolated-preview plugin marketplace.
Ruler4396/dsh-launcher (★9) — Lightweight Windows launcher: silent logon autostart + a minimal WebView2 window instead of a full browser.
bitterSmilezzz/dsh-mac-desktop (★1) — macOS desktop wrapper.
hanelalo/browser-bridge (★17) — Let your agent drive your real browser window like you would.
Lum1104/dsh-browser (★16) — Chrome sidebar extension so DSH operates your browser directly, no vision needed.
whiteguo233/dsh-openbiliclaw (★4) — Bilibili integration for DSH.
vvlife/deepseek-harness-desktop — Self-contained macOS app: bundles the Node runtime + full dsh + full Paseo (daemon + Web UI + mobile pairing) into one APP — drag to Applications and go, zero pre-install. Mobile QR pairing, WhaleHub plugin marketplace, built-in HTML preview, one-click public deploy; isolated from any dsh/Paseo already on your machine.
Zhuchen00123/dsh-wsl-modes (★1) — DSH WSL presets: minimal-wsl + code-wsl with WSL Linux bash + bwrap sandbox and anchored bootstrap.
Vision & Multimodal
中文：让纯文本模型也能"看图"：图像问答、长截图 OCR、UI 还原、像素比对等。
- Anionex/dsh-vision-toolkit (★106) — Vision toolkit for text-only models: intent-aware image Q&A, long-screenshot OCR, UI restoration, grounding, pixel diff, Artifacts, Web UI.
- windyslime/DeepSee (★1) — DSH
0.1.0-rc.5Web-profile vision integration: image turns go through a local DeepSee gateway with pluggable VLM backends while normal text routing stays in DSH. - zhouwumu2-lab/dsh-vision-fix (★10) — Vision fix / repair helper.
- sjscy05/deepseek-harness-vision-plugin — Vision plugin for DSH.
- good-boy4069/Deepseek-omnimodal (★2) — Omnimodal support.
- YYTbit/dsh-plugin-vision-toolkit — Vision-toolkit bridge.
Tools & Editor UX
- Zhangbo-cn/dsh-voice-input-plugin (★6) — Composer mic for the Web UI: tap-to-monitor live transcription and hold-to-talk, with host Edge TTS reply reading that streams while the model generates, echo-pause during reading, and tap-to-stop.
中文：编辑器体验增强、@file 引用、消息分支编辑、会话回滚等"好不好用全靠它"的小工具。
- omdsh-dev/dsh-at-file (★21) — Codex-style
@filementions: search workspace files in the composer and attach their contents to prompts. - omdsh-dev/dsh-custom-tool (★17) — Create & manage sandboxed JavaScript tools with a Monaco editor and model-driven tool lifecycle.
- Moeblack/dsh-message-edit (★9) — Branch-based message editing, reroll, retry, version timeline.
- Anionex/dsh-turn-rewind (★16) — Rewind conversation + workspace state via a persistent Change Ledger.
- Electricitysheep/dsh-tool-turbo (★1) — Tool turbo.
- LingLambda/dsh-undo (★1) — Undo support.
- fakechris/dsh-track (★1) — Tracking helper.
- omdsh-dev/dsh-plugin-skills (★1) — Skills plugin.
- leechen298/Code2Skill (★4) — Generates Function, MCP, Agent Skill, and offline test packages from existing code as an installable DSH bundle.
- omdsh-dev/dsh-mnemon (★1) — Mnemonics plugin.
- ArtificialNotImbecile/dsh-context-taxonomy — Context taxonomy.
- delef/dsh-free-web-search (★0) — Free web search with 10 engines (Bing/DuckDuckGo/SearXNG/AnySearch free + Exa/Tavily/Keenable/Perplexity/DeepSeek paid), automatic fallback chain, time-filtered advanced search, platform search (GitHub/Reddit), web page fetching, LRU caching, and a settings UI. No API keys required.
Agent orchestration & Workflow
中文：多 Agent 团队、可治理的工作流、会话蒸馏等"把一次性调度变成工程资产"的编排类插件。
- NanmiCoder/dsh-agent-teams (★30) — AgentTeams plugin for DSH.
- icetomoyo/dsh_workflow (★29) — Brings Claude Code's UltraCode to DSH; turns one-shot multi-agent dispatch into a generatable / savable / governable / observable / recoverable Workflow layer.
- btspoony/mstar-harness (★38) — Skill-driven Harness / Loop Engineering Workflow Agent Plugin.
- LoserFox/distill (★11) — Automatic conversation distillation: background subagent reflection + skill create/update.
- titanwings/dsh-plannotator (★1) — Plan annotator.
- yyh-001/dsh-companion (★2) — Companion plugin.
- vibeinging/dsh-work (★2) — Work plugin.
- omdsh-dev/dsh-gomoku (★5) — Gomoku game plugin.
- ZK-Andy/dsh-continual-evolve (★11) — Continual self-evolution: versioned, auditable, rollback-safe harness state (prompt notes, memories, skills, subagent specs) refined from session trajectories.
- bpc-oss/dsh-routed-subagent (⭐0) — Run a one-shot subagent fully mounted on any agent preset from any session, with per-call model/provider override, model pre-check, and external CLI engines (codex / claude / codebuddy) with background jobs, live progress, kill, and continuable sessions.
- bpc-oss/dsh-fork-to-preset (⭐0) — Fork any session into a different agent preset from the conversation header: a preset picker creates a new child session mounted on the chosen preset, inheriting the source session's completed turns.
DSH skill bundles (satan9394) — 20 Chinese skill packs：苏格拉底追问 / 热点采集 / 前端设计 / 技能创作 / 合同审查 / 会议纪要 / 求职 / 去AI痕迹 / 工程方法论 / Karpathy 方法论 / PPT / 上下文工程 / 怀疑驱动开发 / 领域建模 / 合并冲突 / 安全加固 / 代码评审 / 会话交接 / 懒人开发 / 学术研究。全部为 skill 型插件（skills/<name>/SKILL.md 经 ctx.skills 注册），每个都是独立 GitHub 仓库并带 dsh-plugin topic：
satan9394/dsh-grill-me (★0) — Socratic grill-me skill: interview until shared understanding via structured questions.
satan9394/dsh-hot-trends (★0) — China real-time hot lists tool+skill (Weibo/Baidu/Bilibili/Zhihu/App Store/QQ Music), no API key.
satan9394/dsh-frontend-design (★0) — Frontend design skill (adapted from Anthropic frontend-design, Apache-2.0).
satan9394/dsh-skill-creator (★1) — Skill creator: author SKILL.md per agentskills.io (adapted from Anthropic skill-creator, Apache-2.0).
satan9394/dsh-contract-review (★0) — Chinese contract risk review with graded reports.
satan9394/dsh-meeting-minutes (★0) — Meeting minutes from transcripts: decisions, action items, owners, deadlines.
satan9394/dsh-career-ops (★0) — Job-search command center: JD A-F scoring, resume optimization, interview prep.
satan9394/dsh-humanizer-zh (★0) — Remove AI-sounding patterns from Chinese text.
satan9394/dsh-superpowers-essentials (★0) — Engineering methodology: classify & approve before coding (Spike/Bounded/Architectural) + systematic debugging.
satan9394/dsh-karpathy-methodology (★0) — Karpathy coding methodology: think first, simplicity, surgical changes, goal-driven verification.
satan9394/dsh-ppt-creator (★0) — Chinese deck generator: outline → visual design → HTML/PPTX output.
satan9394/dsh-context-engineering (★0) — Context engineering: five-level context hierarchy + rules-file discipline.
satan9394/dsh-doubt-driven-dev (★0) — Doubt-driven development: adversarial fresh-context review before non-trivial decisions land.
satan9394/dsh-domain-modeling (★0) — Domain modeling: CONTEXT.md glossary + ADR decision records.
satan9394/dsh-merge-conflicts (★0) — Resolve git merge/rebase conflicts in five steps.
satan9394/dsh-security-hardening (★0) — Security hardening: STRIDE threat modeling + three-tier boundary system.
satan9394/dsh-code-review (★0) — Five-axis code review: correctness/readability/architecture/security/performance.
satan9394/dsh-handoff (★0) — Session handoff document: compact current conversation for the next session/agent.
satan9394/dsh-ponytail-dev (★0) — Lazy senior dev philosophy: YAGNI ladder + root-cause fixes.
satan9394/dsh-academic-research (★0) — Academic research pipeline: question → literature → verification → synthesis → paper → review.
satan9394/dsh-a11y-audit (★0) — WCAG 2.2 accessibility audit: POUR principles, conformance levels, automated + manual verification, remediation guidance.
satan9394/dsh-web-clone (★0) — Web clone: extract fonts/colors/motion/component specs from a reference image/URL and generate a high-fidelity web project.
satan9394/dsh-database-design (★0) — Database table design: PK/normalization/indexes/data types/constraints/performance, PostgreSQL-focused + general rules.
satan9394/dsh-changelog (★0) — Changelog automation: Keep a Changelog + Conventional Commits + semantic versioning, generate release notes from commits/PRs.
satan9394/dsh-postmortem (★0) — Blameless incident postmortems: root-cause analysis (5 Whys), timeline, action items, organizational learning.
satan9394/dsh-runbook (★0) — Incident runbooks: severity levels, detect→triage→mitigate→recover→communicate, escalation paths, on-call handoff.
satan9394/dsh-slo (★0) — SLI/SLO/error budgets: measurable reliability targets, error-budget alerting, SRE practices.
satan9394/dsh-quant-backtest (★0) — Quant backtesting & risk metrics: avoid look-ahead/survivorship bias, transaction costs, walk-forward, Sharpe/drawdown.
satan9394/dsh-api-design (★0) — API & interface design: Hyrum's Law, stable/hard-to-misuse interfaces, REST/GraphQL/module boundaries, change evaluation.
satan9394/dsh-performance (★0) — Performance optimization: measure-first, bottleneck localization, caching/lazy-loading, performance budgets.
satan9394/dsh-git-workflow (★0) — Git workflow & versioning: trunk-based development, commit discipline, semantic versioning, release process.
satan9394/dsh-cicd (★0) — CI/CD automation: quality gates, shift-left, deployment pipelines & strategies, debugging CI failures.
satan9394/dsh-spec-driven (★0) — Spec-driven development: write a structured spec before coding, gated four-phase workflow (spec → approval → implement → verify).
satan9394/dsh-planning (★0) — Planning & task breakdown: verifiable tasks, dependency ordering, estimation, milestones, progress tracking.
satan9394/dsh-deprecation (★0) — Deprecation & migration: code is a liability, safe removal of old systems/APIs, migration lifecycle planning.
satan9394/dsh-observability (★0) — Observability & instrumentation: logs/metrics/traces, structured logging, distributed tracing, alert design.
satan9394/dsh-tdd (★0) — Test-driven development: red-green-refactor loop, Prove-It pattern for bug fixes, tests as proof.
satan9394/dsh-shipping (★0) — Shipping & launch: pre-release checklist, canary/gradual rollout, rollback-first, post-launch monitoring.
satan9394/dsh-idea-refine (★0) — Idea refinement: structured divergent/convergent thinking, turn vague ideas into sharp actionable one-pagers.
satan9394/dsh-incremental (★0) — Incremental implementation: small verifiable steps, keep the system runnable, avoid big-bang changes.
satan9394/dsh-source-driven (★0) — Source-driven development: back every framework decision with official docs, verify and cite sources, no coding from memory.
satan9394/dsh-code-simplify (★0) — Code simplification: remove redundancy, reduce complexity, drop unnecessary abstractions, keep behavior identical.
satan9394/dsh-docs-adr (★0) — Documentation & ADRs: record the why, ADR format, documentation strategy, when not to write docs.
satan9394/dsh-debug-recovery (★0) — Debugging & error recovery: evidence-first, error classification, systematic root-cause, recovery strategies, regression prevention.
satan9394/dsh-browser-testing (★0) — Browser testing & UI verification: verify DOM/console/network/performance in a real browser instead of guessing.
satan9394/dsh-frontend-engineering (★0) — Frontend UI engineering: component design, state management, data fetching, performance & maintainability.
satan9394/dsh-rag (★0) — RAG retrieval-augmented generation: vector databases, embedding, document pipelines, retrieval & citation, hallucination mitigation.
satan9394/dsh-llm-eval (★0) — LLM evaluation: faithfulness/relevance/correctness, test sets, hallucination detection, regression guarding.
satan9394/dsh-data-quality (★0) — Data quality frameworks: validation rules, data contracts, quality monitoring, CI automation.
satan9394/dsh-bash-scripting (★0) — Defensive Bash scripting: set -euo pipefail, argument validation, error handling, debuggability, security.
satan9394/dsh-gitops (★0) — GitOps workflow: declarative infrastructure, Git as source of truth, continuous reconciliation, progressive delivery (ArgoCD/Flux).
satan9394/dsh-pci-compliance (★0) — PCI compliance: cardholder data protection, six-pillar security controls, scope reduction strategies.
satan9394/dsh-agent-teams (★0) — Multi-agent team collaboration: roles, task coordination, parallel workflows, communication protocols.
satan9394/dsh-parallel-dev (★0) — Parallel feature development: file ownership, interface contracts first, vertical slices vs horizontal layers.
satan9394/dsh-prompt-engineering (★0) — Prompt engineering patterns: CoT/ToT, dynamic few-shot, templates, production optimization & debugging.
satan9394/dsh-dataset-curation (★0) — Dataset curation: cleaning, quality filtering, diversity, train/val splits, annotation specs.
satan9394/dsh-data-storytelling (★0) — Data storytelling: SCQA narrative structure, chart selection, title-as-conclusion, credibility.
satan9394/dsh-mlops (★0) — ML pipeline workflow: data→train→evaluate→deploy→monitor, reproducible & regression-guarded MLOps.
satan9394/dsh-microservices (★0) — Microservices architecture patterns: service boundaries, communication, distributed data (Saga), resilience, event-driven.
satan9394/dsh-terraform (★0) — Terraform module library: module design, variable/output conventions, IaC best practices.
satan9394/dsh-architecture (★0) — Architecture patterns: Clean/Hexagonal/DDD tactical patterns, dependency rules, test boundaries.
satan9394/dsh-db-migration (★0) — Database migration: up/down scripts, expand-contract, zero-downtime changes, consistency checks.
satan9394/dsh-auth (★0) — Auth & authorization implementation: JWT/OAuth2/session, RBAC/ABAC, security checklist.
satan9394/dsh-error-handling (★0) — Error handling patterns: layered errors, typed formats, graceful degradation, observability, recovery.
satan9394/dsh-sql-optimization (★0) — SQL optimization patterns: EXPLAIN analysis, indexing strategy, N+1 resolution, query rewriting, slow-query debugging.
satan9394/dsh-monorepo (★0) — Monorepo management: repo structure, workspace dependency management, incremental builds & caching, CI strategy, changeset publishing.
satan9394/dsh-systems-programming (★0) — Systems programming: memory safety patterns (RAII/ownership/smart pointers), concurrency & async patterns, memory/race debugging tools (ASan/Valgrind/Miri/TSan).
satan9394/dsh-startup-business-analyst (★0) — Startup business analysis: TAM/SAM/SOM market sizing, competitive landscape (Five Forces/positioning), financial modeling (runway/cash flow), startup metrics (north star/LTV-CAC).
satan9394/dsh-bazel-build-optimization (★0) — Bazel build optimization: monorepo build config, fine-grained targets, remote caching/execution, pinned deps, visibility governance, build-speed triage.
satan9394/dsh-payment-processing (★0) — Payment processing & subscription billing: billing lifecycle, dunning/proration/tax, Checkout vs PaymentIntent vs SetupIntent, webhook idempotency, PCI compliance, test cards.
satan9394/dsh-event-driven-architecture (★0) — Event-driven architecture: CQRS read/write separation, event sourcing & event store design, projections, Saga distributed transactions with compensation.
satan9394/dsh-service-mesh (★0) — Service mesh: Istio/Linkerd traffic management, mTLS zero-trust, certificate hierarchy & rotation, mesh observability (RED/tracing/access logs).
satan9394/dsh-llm-finetuning (★0) — LLM fine-tuning router & recipes: off-ramps (RAG/prompting), method selection (SFT/DPO/ORPO/KTO/GRPO+RLVR/CPT), LoRA/QLoRA, eval-first, quantized export.
satan9394/dsh-cloud-cost-optimization (★0) — Cloud cost optimization: visibility/tagging/budget alerts, right-sizing, pricing models (RI/Savings Plans/Spot), storage tiering, FinOps cadence.
satan9394/dsh-vector-search (★0) — Vector search engineering: embedding model selection, chunking strategy, index tuning (HNSW/IVF), hybrid search fusion (RRF/rerank), recall evaluation.
satan9394/dsh-blockchain-web3 (★0) — Blockchain & smart contract security: Solidity vulnerability prevention (reentrancy/overflow/access control), CEI pattern, DeFi protocol design, audit prep & attack-path testing.
satan9394/dsh-game-development (★0) — Game development patterns: Godot 4 scenes/signals/state machines & GDScript, Unity ECS (DOTS/Jobs/Burst), performance optimization.
satan9394/dsh-file-conversion (★0) — File format conversion: PDF/Word/HEIC/MP4/CSV/EPUB routes, local tools first (ffmpeg/LibreOffice/Calibre), free online service fallback (999 routes).
satan9394/dsh-signed-audit-trails (★0) — Signed audit trails for agent tool calls: Cedar policy gating (default-deny), Ed25519 hash-chained receipts, offline verification, CI/CD & compliance (EU AI Act/SLSA).
satan9394/dsh-multi-cloud (★0) — Multi-cloud architecture: cross-provider decision framework, service comparison, four patterns (DR/best-of-breed/geo/abstraction), vendor lock-in & data gravity.
satan9394/dsh-python-development (★0) — Python development patterns: src layout & module cohesion, all public APIs, type safety (mypy/pyright), packaging (pyproject/uv), performance & anti-patterns.
satan9394/dsh-hybrid-cloud (★0) — Hybrid cloud networking: on-prem↔cloud connectivity (VPN vs Direct Connect/ExpressRoute), BGP routing & redundancy, active-active DR drills, data residency & gradual migration.
satan9394/dsh-javascript-typescript (★0) — JS/TS development patterns: modern JS best practices, TS advanced types (generics/conditional/mapped/utility), type-safe API clients, Node backend patterns.
satan9394/dsh-social-publishing (★0) — Social media publishing: cross-platform content campaigns, platform-native copy (limits/hashtags/tone), scheduling & rate-limit avoidance, media cropping, analytics review.
satan9394/dsh-frontend-mobile (★0) — Frontend & mobile development: React Native/Expo architecture, Next.js App Router (Server Components/streaming/parallel routes/Server Actions), rendering mode selection.
satan9394/dsh-e2e-testing (★0) — End-to-end testing (Playwright/Cypress): test pyramid boundaries, critical journeys, anti-flaky strategies, CI sharding, failure traces, a11y & responsive checks.
satan9394/dsh-review-agent-governance (★0) — Review-agent governance: human-approval windows gating agent PR reviews/merges/releases/CI edits, Cedar default-deny, Ed25519 receipts, offline chain verification.
satan9394/dsh-x-twitter-research (★0) — X/Twitter research: account research, timeline/trend monitoring, launch-signal tracking, giveaway audits, approval-gated posting (read-first workflow).
satan9394/dsh-codebase-scanner (★0) — Codebase scanner: generate project-doc.md (tech stack/architecture/style/data) & AGENTS.md, full + delta scans detecting drift, context-window-safe analysis.
satan9394/dsh-track-driven-dev (★0) — Track-driven development: tracks as work units (spec.md → plan.md → TDD implementation), phase checkpoints, quality gates, commit SHA records, testable acceptance criteria.
satan9394/dsh-data-engineering (★0) — Data engineering: Airflow DAG design (idempotent/atomic/incremental/observable), dbt layered modeling (staging/intermediate/marts), Spark optimization, pipeline debugging.
satan9394/dsh-kubernetes-operations (★0) — Kubernetes operations: Helm chart scaffolding & templating, Pod Security Standards (restricted), NetworkPolicy & RBAC least-privilege, deploy troubleshooting.
satan9394/dsh-reverse-engineering (★0) — Reverse engineering: binary analysis (disassembly/control-flow patterns), memory forensics, protocol RE, anti-reversing awareness, legal boundaries.
satan9394/dsh-hr-legal-compliance (★0) — HR & legal compliance: employment contract/offer letter/NDA templates, employee vs contractor, GDPR data handling (consent/DSR/privacy by design).
satan9394/dsh-brand-landingpage (★0) — Brand-first landing page design: brand identity interview → design system tokens → generate/feedback loop → deployment-ready HTML (responsive + perf).
satan9394/dsh-framework-migration (★0) — Framework migration: AngularJS→Angular (hybrid/incremental), React modernization, dependency upgrade path (codemod/breaking changes), rollback-safe risk management.
satan9394/dsh-plugin-eval (★0) — Plugin/skill quality evaluation: three layers (static analysis/LLM judge/Monte Carlo), ten scoring dimensions, composite formula, anti-pattern penalties, actionable improvement guide.
satan9394/dsh-bash-testing (★0) — Shell script testing with BATS: error-path & boundary tests with message assertions, fixtures & dependency stubs, CI integration, multi-shell dialect validation.
satan9394/dsh-kpi-dashboard-design (★0) — KPI dashboard design: SMART metric selection, strategic/tactical/operational tiers, dashboard hierarchy (summary/dept/drilldown), visualization best practices, metric governance.
satan9394/dsh-dotnet-backend (★0) — .NET backend patterns: Clean Architecture layering, DI lifetimes, EF Core/Dapper (N+1 prevention), IOptions config, Redis caching, Polly resilience, xUnit testing.
satan9394/dsh-before-you-build (★0) — Pre-build risk review: seven-dimension pre-mortem (demand/positioning/monetization/retention/trust/distribution/adoption), risk verdict, key assumption, smallest validation step.
satan9394/dsh-threat-modeling (★0) — Threat modeling: attack tree construction (OR/AND/leaf with cost/time/skill/detection), STRIDE classification, mitigation mapping & prioritization, residual risk.
satan9394/dsh-operating-kit (★0) — Session operating discipline: start briefing (state doc/live check/drift reconciliation), end finalization (lessons/issues/next steps), stale-state prevention.
satan9394/dsh-content-marketing (★0) — Content marketing: topical clusters & multi-format creation, SEO (SERP gaps/schema/technical), social/email automation, omnichannel distribution, GA4 & CRO review.
satan9394/dsh-observability-tools (★0) — Observability tools: Prometheus config (naming/scraping/recording rules), Grafana dashboards (RED/USE, hierarchy, SLO panels), OpenTelemetry distributed tracing (context propagation/sampling).
satan9394/dsh-distributed-debugging (★0) — Distributed debugging: incident-response flow, metrics/logs/traces correlation, K8s & network troubleshooting, perf profiling, microservices & message-queue issues.
satan9394/dsh-sales-automation (★0) — Sales automation & customer support: personalized cold-email sequences (3-5 touchpoints), follow-up cadence, proposals & objection scripts, empathetic support & closure, conversion tracking.
satan9394/dsh-ai-image-design (★0) — AI image generation workflow: prompt quality rules (50-150 words/self-contained), style guides (photographic/anime/illustration), batch & derivative prompts, brand consistency, review iteration.
satan9394/dsh-functional-programming (★0) — Functional programming: immutability & pure functions, pattern matching, Elixir/OTP (let it crash/supervision), Haskell types (newtype/GADT/total functions), effect isolation.
satan9394/dsh-tech-debt (★0) — Tech debt governance: audit (deps/dead code/markers), interest-risk-cost prioritization, incremental payoff with tests-first, fencing, debt registry & prevention.
satan9394/dsh-api-documentation (★0) — API documentation & developer experience: OpenAPI 3.1, interactive platforms (Swagger UI/Redoc), SDK generation, tested examples, developer portals, lifecycle docs.
satan9394/dsh-full-stack-orchestration (★0) — Full-stack feature orchestration: state-machine driven delivery (state.json), step-by-step with phase checkpoints, per-step artifacts, halt-on-failure, session resume/archive.
satan9394/dsh-deployment-validation (★0) — Deployment validation & config management: config analysis (type/consistency/security), validation schemas in CI, secret scanning, cross-env drift, post-deploy checks & rollback.
satan9394/dsh-web-scripting (★0) — PHP/Ruby web development: PHP 8+ (generators/SPL/type system/OOP), Ruby idioms (blocks/DSL), performance profiling, security & minimal dependencies.
satan9394/dsh-api-scaffolding (★0) — FastAPI project scaffolding: layered structure (api/core/models/schemas/services), async patterns, dependency injection, Pydantic validation, unified error handling, pytest/Alembic/deploy.
satan9394/dsh-security-compliance (★0) — Security audit & compliance: DevSecOps pipeline (SAST/DAST/dependency/container), OAuth2/OIDC/JWT & zero-trust, OWASP & risk management, cloud posture, GDPR/HIPAA/SOC2 mapping, SLSA/SBOM/secrets.
satan9394/dsh-writing-for-agents (★0) — Writing for agents: context pointers & trigger branches, two budgets (context/cognitive load), information hierarchy (steps/reference/progressive disclosure/co-location), completion criteria, leading words, positive phrasing, pruning.
satan9394/dsh-issue-triage (★0) — Issue/PR triage: state machine (bug/enhancement + needs-triage/needs-info/ready-for-agent/ready-for-human/wontfix), verification & grilling, agent-ready briefs.
satan9394/dsh-git-guardrails (★0) — Git safety guardrails: block dangerous commands (push/--force/reset --hard/clean -f/branch -D/checkout .) via PreToolUse hooks, project/global install, human-confirmed overrides, safe alternatives.
satan9394/dsh-teach (★0) — Teaching: stateful learning workspace (MISSION/reference/learning-records/lesson HTML/components), knowledge-skills-wisdom, fluency vs storage strength, desirable difficulty (retrieval/spacing/interleaving), zone of proximal development.
satan9394/dsh-codebase-design (★0) — Deep-module codebase design: small interface + deep implementation, exact vocabulary (module/interface/seam/adapter), interface-as-contract, leverage & locality, testability & AI-navigability.
satan9394/dsh-prototype (★0) — Prototyping: throwaway code that answers a design question (logic/state-model vs UI variants), one-command runnable, no persistence, surfaced state, archive as primary source after validation.
satan9394/dsh-to-questionnaire (★0) — Decision-to-questionnaire: turn unanswered decisions into an async questionnaire for the knowledge holder (grill the send: who/what you need back), gap-targeted questions, theme-grouped most-important-first.
satan9394/dsh-wayfinder (★0) — Wayfinding for huge work: shared decision map on the issue tracker, named destination, decision tickets resolved one at a time, name-based references, index-not-store map, hand off when the path is clear.
satan9394/dsh-setup-wizard (★0) — Interactive setup wizard: bash wizard for human-only steps (provisioning/credentials/CI secrets/third-party dashboards), scoped stages, captured values to .env/secrets, confirmations, ephemeral by default.
satan9394/dsh-workflow-loop (★0) — Delegable workflow design: loop lens for recurring patterns, workflow specs (trigger/checkpoint/push-right/brief), grilled until an implementer can build without questions.
satan9394/dsh-writing-shape (★0) — Writing shape: shape raw material into an article paragraph-by-paragraph, grounding discipline (prerequisites vs introduced concepts), 2-3 candidate openings, write-as-you-go.
satan9394/dsh-writing-beats (★0) — Beat-based writing: assemble material into a choose-your-own-adventure journey, one move per beat, grounding-set-driven reachability, candidate starting/next beats with previews.
satan9394/dsh-writing-fragments (★0) — Writing explore: mine raw fragments (sharp sentences/claims/vignettes/leading words), no structure commitment, capture from the first line, single-file append.
satan9394/dsh-mcp-builder (★0) — MCP server building: four-phase workflow (research & planning, protocol study, framework choice, implement & iterate), tool design (coverage/naming/context/actionable errors), real agent verification.
satan9394/dsh-document-generation (★0) — Office document generation: Word (docx-js gotchas, render verification) & Excel (openpyxl/pandas, zero formula errors, recalc validation, spec-literal), editing existing files by convention.
satan9394/dsh-internal-comms (★0) — Internal communications: 3P updates, company newsletters, FAQ answers, status/leadership/project updates, incident reports (type → guideline → format/tone/gathering).
satan9394/dsh-brand-guidelines (★0) — Brand styling: palette (primary/secondary/accent) & typography pairing (heading/body with fallbacks), background-aware color choice, accent rotation for shapes, hierarchy & readability.
satan9394/dsh-algorithmic-art (★0) — Algorithmic art: two-step flow (create an algorithmic philosophy → express in p5.js), seeded randomness/noise fields/particle systems, interactive parameter exploration, originality discipline.
satan9394/dsh-theme-factory (★0) — Theme factory: 10 preset professional themes (palette + font pairing) for slides/docs/HTML, showcase→choose→confirm→apply, on-the-fly theme generation, contrast & readability.
satan9394/dsh-doc-coauthoring (★0) — Doc co-authoring: three-stage workflow (context gathering with meta questions/templates/info-dump → refinement & structure → reader testing with a fresh no-context session), alt-text checks.
satan9394/dsh-discernment-nudge (★0) — Discernment nudge: after substantive answers, 2-3 content-bound follow-ups (check facts / question reasoning / notice missing context), at most once per conversation, with skip conditions.
satan9394/dsh-gif-creator (★0) — GIF creation: search via GIF APIs (keyword/licensing), custom text/frame animations, size/loop optimization, platform-fit (Slack) & share-ready output.
satan9394/dsh-llm-api-integration (★0) — LLM API integration patterns: SDK vs raw HTTP (no mixing), streaming default, thinking modes, tool use, prompt caching, drift verification (never guess SDK usage), model migration & prompt audit.
satan9394/dsh-prompt-audit (★0) — Prompt audit & model migration: scan prompts/skills/tool descriptions for stale patterns, scope & target model, inventory & provenance, report (file:line/reason/confidence) + suggested diff.
satan9394/dsh-recsys-pipeline (★0) — Recommendation/ranking pipeline: six-stage framework (Source→Hydrator→Filter→Scorer→Selector→SideEffect) for any top-K problem (feeds, search ranking, RAG rerankers, task prioritizers).
satan9394/dsh-webapp-testing (★0) — Web app testing with Playwright: decision tree (static HTML vs dynamic server), recon-then-act (networkidle → screenshot/DOM → selectors → actions), server lifecycle helper, browser logs.
satan9394/dsh-sast-security (★0) — SAST tooling: Semgrep/SonarQube/CodeQL config, custom rules, CI quality gates, false-positive tuning, scan performance, defense-in-depth with DAST/dependency scanning.
satan9394/dsh-security-requirements (★0) — Security requirement extraction: three-layer translation (business → security requirements → technical controls), types (functional/non-functional/constraint), traceability/testability/priority, user stories & acceptance criteria.
satan9394/dsh-pdf-processing (★0) — PDF processing: read/extract text & tables, merge/split, rotate/watermark, create, fill forms, encrypt/decrypt, extract images, OCR (pypdf/pdfplumber/reportlab).
satan9394/dsh-canvas-design (★0) — Canvas visual design: create a design philosophy (form/space/color/composition) then express it as poster/art (90% visual + 10% text), original work only.
satan9394/dsh-commit-message (★0) — Git commit message conventions: Conventional Commits (type(scope): subject), body/footers, BREAKING CHANGE, issue refs, atomic commits, semver linkage.
satan9394/dsh-web-artifacts (★0) — Interactive web artifact building: React 18+TS+Tailwind+shadcn/ui scaffold → develop → bundle single HTML, anti-AI-slop design guidelines, responsive & accessible.
satan9394/dsh-pptx-engineering (★0) — PPTX engineering: pptxgenjs gotchas (layout/hex colors/shadow/lists/native charts), unzip-edit-zip editing, markitdown/thumbnail reading, schema validation & render checks.
satan9394/dsh-css-art-styles (★0) — CSS art style recipes: reusable visual systems (liquid-glass variants, cinematic space) with design tokens, utility CSS, entrance animations, layout skeletons.
satan9394/dsh-diagram-design (★0) — Editorial diagram design: 38 diagram types as self-contained HTML/SVG (architecture/flow/sequence/ER/swimlane/Wardley/Sankey/fishbone), deletion-first philosophy, brand style gate, semantic patterns.
satan9394/dsh-screenshot-to-code (★0) — Screenshot-to-code: analyze layout/components/tokens → generate HTML/Tailwind/React → render-and-compare iteration (pixel alignment/responsive/interaction) → deliver.
satan9394/dsh-agent-loop-engineering (★0) — Agent loop engineering: L1 auto/L2 assisted loop tiers, loop patterns (daily triage / PR babysitter / dependency sweeper / changelog drafter), multi-loop coordination, verifier + human gates, loop-audit (readiness/cost), anti-patterns.
satan9394/dsh-cad-modeling (★0) — Parametric CAD modeling: STEP-first (build123d Python source or direct STEP import), natural-language specs / reference images / 2D drawings, assemblies with source-level joints & mating datums, geometry checks, STL/3MF/GLB exports.
satan9394/dsh-cli-anything (★0) — Make any software agent-native: 7-phase CLI harness pipeline (analyze→design→implement→test→publish), real-backend integration (no toy implementations), JSON output + REPL dual mode, refine gap analysis, output verification (never trust exit 0).
satan9394/dsh-personal-content-discovery (★0) — Cross-platform personalized content discovery: five-layer soul portrait (event→preference→awareness→insight→soul), MBTI inference, active exploration breaking filter bubbles, cross-platform interest fusion, local-first, friend-style explainable recommendations.
satan9394/dsh-colleague-creation (★0) — Distill a real person into an AI skill: source materials (chat/docs/email/screenshots) + subjective description → a skill that works like them (Work Skill + 5-layer persona, personality/company-culture/level tags, incremental evolution, version rollback).
satan9394/dsh-designmd (★0) — DESIGN.md design-system format for coding agents: YAML design tokens (colors/typography/rounded/spacing/components) + Markdown rationale, token references, lint/diff/export toolchain (Tailwind v3/v4, W3C DTCG), WCAG AA contrast checks.
satan9394/dsh-content-distillation (★0) — Distill books/videos/podcasts into callable AI skills: RIA-TV++ 7-phase pipeline (Adler analysis → 5 parallel extractors → triple verification 25-50% pass → RIA++ construction → Zettelkasten links → trap-question stress tests → DIGEST.md + installable skills).
satan9394/dsh-investment-research (★0) — Personal investment research methodology (A-share/US/HK): five-dimension analysis (valuation/capital/financials/industry/catalyst-risk), bull-bear debate (fact sheet first → bull/bear → neutral moderator lists divergence + verification checklist), reflection audit, daily review — deliberately never outputs buy/sell calls.
satan9394/dsh-relationship-coach (★0) — Relationship/dating coach: emotions first then solutions, fact/guess/unknown separation, reciprocity/reality/risk/opportunity-cost comparison, sendable scripts + observation windows + stop conditions, safety-first for dangerous situations.
satan9394/dsh-frontend-slides (★0) — Zero-dependency HTML presentations: show-don't-tell visual style discovery (3 previews to pick from), PPT→web conversion preserving assets, anti-AI-slop curated styles, 34 bold design systems with progressive loading (index → preview cards → full design.md only when picked).
satan9394/dsh-html-ppt (★0) — HTML PPT Studio: 36 themes (token-driven reskin) × 31 layouts × 47 animations (27 CSS + 20 canvas FX) × 15 full-deck templates, presenter mode (S-key magnetic cards + speaker script + timer, BroadcastChannel-synced), keyboard navigation, zero build step.
satan9394/dsh-brand-design (★0) — Brand design system: 5-step brand asset protocol (ask → search official brand pages → download assets → grep hex colors, never guess → freeze brand-spec.md), direction advisor (3 parallel logic subagents → 3 real visual options), Junior Designer workflow, anti-AI-slop rules, 5-dimension expert review.
satan9394/dsh-caveman-speak (★0) — Terse token-saving output discipline: caveman-style answers (~65% fewer output tokens) while code/commands/errors stay byte-exact; token-sink audit (safe fix / offload / habit / load-bearing classes), progressive-disclosure recall.
satan9394/dsh-persistent-memory (★0) — Cross-session persistent memory compression: lifecycle-hook observation capture, 3-layer retrieval workflow (search index → timeline → full details, ~10x token savings), progressive disclosure, private tags, hybrid vector+keyword search.
satan9394/dsh-planning-files (★0) — Persistent 3-file planning: task_plan.md / findings.md / progress.md on disk re-injected every turn, plan survives /clear, crashes and compaction; session catchup recovery, isolated parallel-task dirs, opt-in completion gate.
satan9394/dsh-skill-seekers (★0) — Knowledge data layer for AI systems: 18 source types (docs sites, GitHub repos, PDFs, videos, notebooks, wikis, chat exports) → structured assets, export to 22 targets (Claude/Gemini/OpenAI skills, RAG pipelines, coding assistants), AI-driven project scanning.
satan9394/dsh-scientific-research (★0) — Scientific skill pack: 100+ provenance-rich database lookups (PubChem/UniProt/ClinicalTrials.gov/FRED…), literature review + evidence-bounded hypothesis generation, costed experimental protocols, point-by-point peer-review rebuttals, method validation (ICH/USP/CLSI), publication-grade figures.
satan9394/dsh-marketing-growth (★0) — Marketing growth skill tree: product-marketing foundation read first by all skills, seven clusters (SEO/CRO/copy/ads/retention/GTM/strategy), skill cross-references, marketing loops (recurring auto-run workflows), measurable experiment design.
satan9394/dsh-obsidian-vault (★0) — Obsidian vault operations: Obsidian Flavored Markdown (wikilinks/embeds/callouts/properties), Bases views (filters/formulas/summaries), JSON Canvas (nodes/edges/groups), Obsidian CLI, Defuddle clean-Markdown web extraction to save tokens.
satan9394/dsh-live-docs (★0) — Live docs injection against hallucination: pull version-specific up-to-date docs before writing library/API code (CLI + MCP dual mode), library-ID direct fetch, explicit version pinning, rule-based auto-trigger — kills hallucinated APIs and stale examples.
satan9394/dsh-last30days (★0) — Cross-platform last-30-days research: 16+ sources (Reddit/X/YouTube/TikTok/HN/Polymarket/GitHub/arXiv) searched in parallel, scored by real engagement (upvotes/likes/real money), AI judge synthesizes one brief — pre-meeting people research, hiring signals, discovery of pre-peak topics, tool comparisons.
satan9394/dsh-repo-graphify (★0) — Repository knowledge graph: multimodal extraction (AST/LLM/vision) of concepts & relationships, honest EXTRACTED/INFERRED/AMBIGUOUS edge tags, god nodes & surprising connections, ~71.5x token savings, wiki navigation, --watch auto-sync, git-commit hook rebuild.
satan9394/dsh-adhd-friendly (★0) — ADHD-friendly action-first outputs: lead with next action, numbered steps, end with one concrete next step, suppress tangents, restate state, specific time estimates, visible wins, matter-of-fact errors, lists capped at 5, no preamble/recap/closers.
satan9394/dsh-better-interface (★0) — Interface quality skill series: six dimensions (UI polish/typography/colors/accessibility/layout/UX writing), variant (multiple genuinely different versions behind a picker, promote the winner), explain-interface (reverse-engineer any effect), interface-review (full-dimension review of changes/branch/PR).
satan9394/dsh-opinion-analysis (★0) — Multi-agent opinion analysis: 30+ domestic/global social platforms & millions of comments, four specialist agents (Query/Media/Insight/Report) + debate moderator, agent-forum chained deliberation against homogeneity, multimodal short-video parsing, public+private data fusion, interactive HTML report.
satan9394/dsh-office-cli (★0) — Office suite designed for AI agents: single binary controls Word/Excel/PowerPoint (no Office install), path selectors (/slide[1]/shape[1]) for precise targeting, HTML rendering engine gives AI eyes (render→look→fix loop), live preview real-time feedback, outline/HTML/JSON views.
satan9394/dsh-swarm-prediction (★0) — Swarm intelligence prediction: build a high-fidelity parallel digital world from seed information, thousands of agents with independent personas/memory/behavior freely interact and socially evolve, inject variables god's-eye view to deduce future trajectories — rehearse the future in a digital sandbox.
satan9394/dsh-hallmark-design (★0) — Anti-AI-slop UI design: 21 themes + macrostructure per brief, 57 slop-test gates + pre-emit self-critique, refuses on-distribution LLM defaults; four verbs (build / audit / redesign / study-DNA), Custom from-scratch mode.
satan9394/dsh-agent-reach (★0) — One-click internet capability for AI agents: 15+ platforms (web/YouTube subtitles/RSS/search/GitHub/X/Bilibili/Reddit/FB/IG/Xiaohongshu/LinkedIn/V2EX/Xueqiu/podcasts), zero-config first, multi-backend routing auto-failover, doctor diagnostics, cookies stay local.
satan9394/dsh-harness-os (★0) — Agent harness config OS: four-layer separation (rules persistent / skills on-demand / agents specialized / hooks runtime), instincts continuous learning (+/evolve clustering), AgentShield security audit (secrets/hook-injection/MCP risk), Unified Memory Vault across harnesses, GateGuard destructive-command gate.
satan9394/dsh-self-improving-agent (★0) — Self-improving AI agent: built-in closed learning loop (auto-create skills from experience, improve during use, FTS5 session search + LLM summaries for cross-session recall, Honcho dialectic user modeling), cross-platform gateway (Telegram/Discord/Slack/WhatsApp), cron natural-language scheduling, subagent parallelism, serverless hibernate/wake.
satan9394/dsh-social-simulation (★0) — Million-agent social simulation: scalable open-source social-media simulator where LLM agents mimic up to 1M users on Twitter/Reddit-style platforms, 23 actions, interest/hot-score recommendation, dynamic environments — study information spread, group polarization, herd behavior.
satan9394/dsh-autonomous-research (★0) — Autonomous LLM training research: give an agent a small real training setup to experiment overnight (edit code → fixed 5-min train → check val_bpb → keep/discard loop, ~12 experiments/hour), single-file modification, program.md as ultra-lightweight skill.
satan9394/dsh-cloud-well-architected (★0) — Cloud Well-Architected six-pillar review: reliability (user-experience goals, SLO + error budgets, cross-region redundancy, horizontal scaling, golden signals, graceful degradation, recovery tests, blameless postmortems), cost, operational excellence, performance, security, sustainability.
satan9394/dsh-skill-optimization (★0) — Train agent skills like neural networks: epochs/batchsize/learning-rate/validation gates without touching model weights, rollout→reflect→aggregate→select→update→evaluate loop, candidate edits accepted only on strict held-out improvement, zero inference-time overhead, deploy compact best_skill.md.
satan9394/dsh-geo-seo (★0) — GEO-first AI search optimization: optimize sites for AI engines (ChatGPT/Claude/Perplexity/Gemini/AI Overviews), citability scoring, AI crawler analysis, llms.txt generation, brand-mention scanning, Schema + E-E-A-T, 5 parallel subagent full audit.
satan9394/dsh-virtual-eng-team (★0) — Virtual engineering team: 23 specialist roles (CEO rethinks product, eng manager locks architecture, designer catches AI slop, reviewer finds prod bugs, QA opens a real browser, security officer runs OWASP+STRIDE, release engineer ships PRs), slash-command flow /office-hours → /autoplan → /review → /qa → /ship.
satan9394/dsh-knowledge-brain (★0) — Knowledge brain layer: search gives pages, brain gives cited answers — synthesis layer + self-wiring knowledge graph (zero-LLM entity edges) + gap analysis (what the brain doesn't know) + 24/7 dream cycle consolidation + company-brain per-login isolation.
satan9394/dsh-channel-assistant (★0) — Channel-first personal assistant: one Gateway connects models/tools/messaging channels/companion apps, channels-as-interface (WhatsApp/Telegram/Slack/Discord/Signal/iMessage), Control UI/CLI/TUI, pairing-based security for unknown senders, sandboxed tool execution, ClawHub plugin ecosystem.
satan9394/dsh-desktop-agent-gui (★0) — Desktop GUI for agent orchestration: zero-config wizard, intelligent chat (@agent routing, inline skill cards, doc previews), multi-channel multi-account management, cron automation delivering to external channels, local-first skill management, keychain-secured provider credentials.
satan9394/dsh-model-gateway (★0) — Model routing gateway: one endpoint to 290+ AI providers (90+ free, 1200+ models), Claude Code/Codex/Cursor/Cline/Copilot compatible, quota-aware auto-fallback, 19 routing strategies, RTK+Caveman stacked compression saving 15-95% tokens, honest free-tier budget (~1.5B tokens/mo).
satan9394/dsh-parallel-agent-ade (★0) — Parallel agent development environment: Codex/ClaudeCode/OpenCode/Pi side-by-side each in its own git worktree, fan one prompt across agents and merge the winner, mobile companion monitoring, Design Mode (click UI element → HTML/CSS/screenshot into prompt), SSH remote worktrees, annotate-AI-diff loop.
satan9394/dsh-doc-compiled-skills (★0) — Precompile official docs into executable skills: extract task-oriented capabilities → classify (limits/troubleshooting/config/architecture/security) → reveal on user intent; knowledge pre-compiled into actions/choices/guardrails instead of per-query RAG retrieval.
satan9394/dsh-html-template-library (★0) — Reusable HTML template library methodology: 34 design systems (Soft Editorial / Neo-Grid Bold / Stencil & Tablet / Sakura Chroma / Vellum / Monochrome…), agent reads index.json, matches the brief, clones and adapts — 3 multi-layout previews + metadata + reusable design tokens per template.
satan9394/dsh-taste-review (★0) — Taste-driven judgment call for ambiguous decisions: state the question plainly, list files to consider, weigh a few options, give recommendation with alternatives; depth matches the decision weight — design calls may warrant several paragraphs, naming calls may not.
satan9394/dsh-test-desktop-app (★0) — Electron desktop app E2E testing via CDP: connect to real Chromium renderer for DOM inspection, JS evaluation, clicks, and screenshots; DPR×2 multiplier, real Input.dispatchMouseEvent for right-clicks, stacking checks for popup screenshots, debounced xterm refit.
nortejiang-tech/dsh-req-miner (★0) — Requirements-mining sidebar plugin: per-session floating interview window driven by a continuable subagent (decision tree + frontier questions), reads the bound session's workspace and recent context, one-click return of the summarized requirement prompt to the composer. Install:
github:nortejiang-tech/dsh-req-miner.
Integrations & Bridges
中文：把 DSH 接到 VS Code、桌面通知、或其它 Agent（Claude / Codex / Pi / OpenCode）的桥接类插件。
PandaPolo/dsh-voice-call (★1) — Agent-initiated voice calls:
offer_callrings the human (接听/拒接/稍后再说), accepted calls synthesize and play locally via CrispASR + Qwen3-TTS (9 speakers, 2 Chinese dialects), rejected calls return the decision to the agent.omdsh-dev/dsh-open-in-vscode (★28) — Open workspace directories in VS Code directly from the web GUI.
omdsh-dev/dsh-notification (★19) — Desktop notifications for turn completions, with per-outcome controls and include/exclude keyword rules.
Nagi-ovo/dsh-find-plugins (★12) — In-app plugin finder.
YYTbit/dsh-plugin-claude-bridge — Bridge to Claude.
YYTbit/dsh-plugin-codex-bridge — Bridge to Codex.
YYTbit/dsh-plugin-pi-bridge — Bridge to Pi.
YYTbit/dsh-plugin-opencode-bridge — Bridge to OpenCode.
bobleer/deepseek-harness-plugin-mcp — MCP plugin.
labmimors/dsh-mcp-lens (★5) — Progressive-disclosure MCP gateway for DSH: search remote tools, inspect exact input schemas on demand, then call an explicit server/tool pair.
yoke233/dsh-openai-codex-auth (★1) — OpenAI Codex auth.
vvlife/dsh-agnes-paseo — Agnes AI model gateway (OpenAI-compatible) for dsh, plus a zero-dependency ACP bridge that registers DeepSeek Harness as a Paseo provider.
nickhelion/dsh-qwen-token-plan-cn-responses (★0) — Qwen Token Plan CN Personal Responses API provider: official model/tool catalog sync, server-side web/code/extraction/image tools, DSH local functions, and image input.
nickhelion/dsh-serverchan-notify (★0) — ServerChan3 (Server酱) push notifications when an agent turn finishes — title, model, project directory, git branch, status and a reply excerpt, with subagent filtering and fire-and-forget delivery.
vvlife/dsh-paseo-mobile — Connect your phone to dsh via Paseo: one-command setup registers dsh as a Paseo provider (zero-dependency ACP bridge), then scan the pairing QR in the Paseo mobile app. Model-agnostic: follows your existing dsh model config; mirrors dsh web sessions to the phone with context-aware follow-ups.
SwainGao/dsh-plugin-ai-bridge (★1) — Bridge to external AI models (Codex / Claude / GPT / OpenAI-compatible relays) for read-only second-opinion code review, adversarial review, task delegation with resume threads, and non-blocking background jobs.
dsh plugin add dsh-plugin-ai-bridge@0.1.3.Nwflower/dsh-chat-import (★54) — Import Claude Code / Codex / ChatGPT / Cursor chat histories as resumable DeepSeek Harness sessions.
Sidebar, Workspace & Ecosystem
中文：侧边栏工作台、oh-my-dsh 这类"插件库"、插件脚手架与注册表等生态基础设施。
- omdsh-dev/DSH-better-sidebar (★66) — Full workbench sidebar with third-party tab registration: file render/edit, terminal, Git, subagent.
- sandbaseai/dsh-plugin-store (★1) — Native Store for DSH Settings: live catalog search, tag filters, ranking, runtime-verified installation, and installed-plugin inventory.
- LaplaceYoung/oh-my-dsh (★12) — Plugin ecosystem: 700+ plugins wired only through extension seams, no agent-loop changes.
- kingjly/dsh-plugin-builder (★1) — Plugin builder scaffolding.
- vlln/plugin-registry (★6) — Plugin registry.
- DeKrych/Dshell-plugins (★27) — Dshell plugin collection.
- HackSing/dsh-plugins / Yihong89/dsh-plugins — Plugin collections.
- coppynight/dsh-doctor (★2) — Diagnostics / doctor.
- yyh-001/dsh-expression (★1) — Expression plugin.
- Chinesezjc/dsh-interconnect (★8) — Cross-instance message/event handoff.
Fun & Misc
中文：合影墙、音乐、框架类实验等"好玩 / 杂项"插件。
- SenmuuuuW/dsh-group-photo (★11) — Internal-test group-photo wall (GitHub OAuth, frozen allowlist).
- syy-shark/dsh-music-plugin — Music plugin.
- unknowbug/RE-Framework (★5) / unknowbug/anchorlaw (★4) — Frameworks.
- hxs996-beep/deepAct (★7) — deepAct.
- aga-j/dsh-mini-games — Pure-frontend mini-game collection in the web details panel: guess-the-number, 2048, minesweeper (
dsh plugin --profile web add dsh-mini-games).
Hands-on Notes
中文 · 实战评测：下面挑了 6 个有代表性的高星插件，按"怎么装 / 怎么用 / 坑点"写成可直接照着做的短评测（star 数为 2026-08-13 当晚撰写时数据）。涉及 dsh-external/* 私有仓库的，已标注需要读取权限。
English · hands-on notes: short, copy-pasteable reviews of six representative
high-star plugins — install / use / gotchas. Star counts are from the night of
2026-08-13. Entries under dsh-external/* are private repos and need read
access.
dsh-web-ui — Web UI 全家桶（★311）
装：npm 已发布到 @linxin666 scope，推荐直接装聚合包：
dsh plugin --profile web add @linxin666/dsh-web-ui-all；只要皮肤就装
@linxin666/dsh-skins。装完重启 dsh web，侧边栏即出现全部入口。
用：任务看板（支持 cron 定时让 DSH 会话自动执行，如每天升级 DSH / 周一生成周报）、 Git 图谱、右侧预览面板（Markdown/HTML/代码/diff/CSV/PDF/Office/图片）、鲸鱼娘宠物、 实时 TPS 与令牌统计、移动端远程控制、以及 SSH 远程运维（xterm 终端 / SFTP / 端口转发 / 集群并发执行）。
坑：首次安装若报 ERR_PNPM_IGNORED_BUILDS，按提示把 cloudflared / ssh2 等加入
profile 的 pnpm-workspace.yaml 的 allowBuilds 再重跑。验证是否挂上可用
dsh --profile web --dump-config。
Install @linxin666/dsh-web-ui-all to get the whole bundle (task board, git
graph, right panel, pet, live token stats, mobile remote, SSH ops). Add
@linxin666/dsh-skins for skins only. Restart dsh web after install. If you
hit ERR_PNPM_IGNORED_BUILDS, allowlist cloudflared/ssh2 in the profile's
allowBuilds.
dsh-cc-tui — Claude Code 风格全屏终端（★103）
装：dsh plugin --profile cc-tui add dsh-cc-tui（会自动初始化 cc-tui
profile），然后 dsh --profile cc-tui 启动；或仓库根目录 sh install.sh
（Windows 用 dsh-cc.cmd，支持 --resume 恢复上次会话）。
用：像素鲸鱼顶栏 + 启动手绘动画、实时工作状态行（在跑哪个工具 / 思考文案）、
思考过程流式展开、双击 Esc 时间回溯（fork 重放历史消息）、底部蓝白上下文进度条 +
TPS 仪表、复刻 CC 的 / 命令菜单（/plan /goal /compact /review 等全部走官方链路）。
坑：需要官方 dsh CLI 与 pnpm；纯插件挂载，卸载即完全还原。它不消费审批流
（/permissions 仅说明现状）；/model 实时切换走"会话 fork 续聊"而非原位换模型。
A Claude-Code-style full-screen TUI: pixel-whale banner, live status row,
streaming thinking, double-Esc rewind (fork + replay), context bar + TPS meter,
and CC-style / commands that all ride DSH's official services. Pure Cordis
mount — removing the plugin fully restores the host. No approval-flow UI yet.
DSH-better-sidebar — 侧边栏完整工作台（★71）
装：npm 已发布 dsh-better-sidebar@0.10.2：
dsh plugin --profile web add dsh-better-sidebar@0.10.2，装完重启 DSH 并硬刷新
（Cmd/Ctrl+Shift+R）。嫌麻烦可 curl -fsSL <repo>/scripts/install.sh | bash，脚本会
自动处理 allowBuilds 与旧挂载行。
用：右侧栏 + 底部面板双工作台——资源管理器（懒加载目录树、@文件 引用）、
CodeMirror 编辑预览（Office/PDF/HTML 沙箱 iframe）、内嵌浏览器（沙箱 iframe）、
xterm 真实终端（每会话最多 3 个）、Git 面板（真 diff + 暂存/提交）、后台任务页。
暴露 ctx.betterSidebar 服务，其它插件可注册侧边栏 Tab 与文件预览器。
坑：pnpm 11 的 strict-dep-builds 可能拦截 node-pty 构建，执行
pnpm approve-builds --all 再重跑；Git 面板无 push/pull/fetch；
.xlsx 不保留单元格样式，Office 预览约 23MB 首次较慢。
Right-sidebar + bottom-panel workbench: file explorer, CodeMirror editor with
Office/PDF/HTML preview, sandboxed in-app browser, real xterm terminal, git
panel, background-task view. Exposes ctx.betterSidebar so other plugins can
register tabs/viewers. On pnpm 11 run pnpm approve-builds --all if node-pty's
build is ignored. Git panel has no push/pull; Office preview is ~23MB on first load.
dsh-vision-toolkit — 让纯文本模型"看见"（★110，私有仓库）
装：目前是 dsh-external 私有 GitHub release，需对该仓库的读取权限。
git clone https://github.com/dsh-external/dsh-vision-toolkit.git 后
dsh plugin --profile web add "$PWD/dsh-vision-toolkit"（Web 和 Headless 各加一次）。
Web 端进 Settings → Vision Toolkit，配一个 DSH Credential（如 VISION_API_KEY）
并点 Test connection。
用：会话里把图片放进 workspace 路径，调用 /vision-tools 激活，再让 Agent 用
vision_glance（图像问答）、vision_ground/vision_detect（定位框）、
vision_crop/vision_trace（裁剪/转 SVG）、vision_pixel_diff（像素比对）、
vision_long_screenshot_ocr（长截图 OCR）等 10 个工具。本地裁剪 / trace / 像素差
不需要视觉 API；远程识别需一个 OpenAI 兼容视觉端点（managed runtime 会自动装好
Python 3.11+ 环境）。
坑：私有仓库，没权限 git clone 会失败；远程工具需要配 Credential，否则只有本地
工具可用。
Gives text-only DSH agents eyes via 10 structured vision tools (Q&A, grounding,
OCR, pixel diff, UI restoration…). Private dsh-external repo — needs read
access. Local crop/trace/diff need no vision API; remote recognition needs an
OpenAI-compatible endpoint configured as a DSH Credential. Managed runtime
auto-prepares Python 3.11+.
oh-my-dsh — 687 个插件的能力库（23 轮差距登记，5286 测试全绿）
定位：从 opencode / oh-my-pi / Codex / Claude Code / pi / Goose 等对照，把有用能力
以 DSH 插件形态重写——只走扩展接缝（ctx.effect() / ctx.on() / ctx.tools），
不改 agent-loop、不引入热路径开销。已登记 23 轮差距、687 个插件。
用：仓库按 plugins/<gap-id>/ 组织，每个都是独立 DSH 插件。
pnpm install && pnpm test 跑全部；单个插件 cd plugins/<gap-id> && pnpm test。
端到端验证（e2e/run-e2e.sh）需要 DSH_HOME/.env 里的 DEEPSEEK_API_KEY。
坑：这是插件"源码库 / 能力库"，不是一条 dsh plugin add 就能装全家桶的聚合包；
要按需挑选具体插件单独安装。
A capability library: 687 plugins rewritten from opencode / Claude Code / Codex /
pi / Goose etc., all through DSH's extension seams (no agent-loop changes). 5286
tests green. It's a source/ability repo, not a one-line aggregate — pick and
install individual plugins under plugins/<gap-id>/. e2e needs DEEPSEEK_API_KEY.
dsh_workflow — 把多 Agent 调度升级成可治理 Workflow（★30，私有仓库）
装：dsh-external 私有仓库，
dsh plugin --profile web add "github:dsh-external/dsh_workflow#main"，
dsh --profile web --dump-config 验证配置里出现 dsh-external-workflow 行后重启。
用：会话里 /workflow list 看内置/项目/个人 workflow；/workflow parallel-investigation {"question":"…"} 跑并行调研；/workflow create 设计一个并行安全评审流程 现场生成；
/workflow review --risk high --requirement "不得破坏公开 API" --wait 跑受控评审。
支持命名/项目/个人 workflow、run graph、暂停/恢复/重跑/续跑、成本记录，以及
capability-only VM（QuickJS WASM）安全边界。
坑：需 Node >=22.19 且与 compatibility.json 一致的 DSH 快照；github: 安装需对该
私有仓库读取权限；/workflow create 形式不接受 --wait，同步等待请用命名 workflow
的 --wait。同类多 Agent 主题另见 dsh-agent-teams。
Upgrades one-shot multi-agent dispatch into a governable Workflow layer:
named/project/personal workflows, run graph, pause/resume/rerun, cost records,
and a QuickJS-WASM capability-only sandbox. Private dsh-external repo —
github: install needs read access and Node >=22.19. /workflow create does not
accept --wait; use it on named workflows. For teams of agents, see
dsh-agent-teams.
Other awesome lists (meta)
These are community "awesome" indexes for DSH — useful cross-references, some with daily compatibility tracking:
- AdamPlatin123/awesome-dsh-plugins (★187) — Directory with daily compatibility tracking.
- 0xsline/awesome-deepseek-harness (★84) — Curated plugins/tools/infra from
dsh-external/huband thedsh-plugintopic. - Alex-Yanggg/awesome-DSH-plugin (★27)
- awesome-dsh-plugin/awesome-dsh-plugin (★19)
- bruc3van/awesome-dsh-plugin (★8)
- walkinglabs/awesome-deepseek-harness-plugins (★1)
Contributing
中文：两种方式让插件被收录，二选一即可——
通道 A · 零维护（推荐）：直接给你的 GitHub 仓库打上 dsh-plugin 话题。
PLUGINS.md 由 update.yml 每天从
该话题自动抓取刷新，无需提 PR，约 24 小时内就会出现在快照里。
通道 B · 进精选列表：想进上面人工分类的精选 README？提一个 PR，把它加到合适的 分类（附一句话描述与当时的 star 数）。这个 PR 会被 review-pr.yml 按规则自动审核：
- ✅ 只能改
README.md（不能顺手改其他文件）； - ✅ 新增的每个仓库必须公开存在；
- ✅ 必须明显是 DSH 插件——要么带
dsh-plugin话题，要么名称/描述里出现 DeepSeek Harness / DSH / Cordis / 外挂 / 插件 等字样； - ✅ 不能与已有条目重复；
- ⚠️
dsh plugin add <pkg>安装命令需格式正确。
全部通过会打 auto-approved 标签并自动合并；不通过会打 changes-requested
并指出原因，改完重推即可。维护者仍可随时人工覆盖。
完整规则见 CONTRIBUTING.md。
Found or built a plugin? Make it discoverable — pick either:
Channel A · Zero maintenance (recommended): just add the dsh-plugin
topic to your GitHub repo. PLUGINS.md is fetched daily from that
topic by update.yml — no PR needed; it appears in
the snapshot within ~24h.
Channel B · Curated list: want it in the categorized README above? Open a PR adding it to the right section (one-line description + star snapshot). The PR is auto-reviewed by rules via review-pr.yml:
- ✅ Only
README.mdmay be changed. - ✅ Every added repo must exist and be public.
- ✅ Must be clearly DSH-related: carry the
dsh-plugintopic, or mention DeepSeek Harness / DSH / Cordis / 外挂 / 插件 in name or description. - ✅ No duplicates of existing entries.
- ⚠️
dsh plugin add <pkg>install commands must be well formed.
All checks pass → auto-labeled auto-approved and auto-merged. Otherwise
changes-requested with reasons; push a fix to re-run. A maintainer can still
override at any time.
Full rules: CONTRIBUTING.md.
License
The list content is released under CC0 1.0. Individual plugins keep their own licenses (mostly MIT, some CC BY-NC-SA for skins).
相关推荐Related
-
ruvnet/ruflo
🌊 The original agent meta-harness. Deploy intelligent multi-player swarms, coordinate autonomous workflows, and build c…
-
Tencent/WeKnora
Open-source LLM knowledge platform: turn raw documents into a queryable RAG, an autonomous reasoning agent, and a self-m…
- Nagi-ovo/voyager Enhancement suite for Gemini, AI Studio, Claude & ChatGPT — plus a prompt manager for any web UI, DeepSeek Harness inclu…
-
EverMind-AI/EverOS
One portable memory layer for every AI agent: local-first, Markdown-native, user-owned, and self-evolving across apps, t…
-
freestylefly/awesome-gpt-image-2
Prompt as Code | GPT-Image2 工业级提示词引擎与模板库，470+ 个案例逆向工程，20+ 套工业级模板，并提炼出Skills，持续更新中
-
awesome-dsh-plugin/awesome-dsh-plugin
A curated list of plugins for DeepSeek Harness (dsh) · DeepSeek Harness 插件精选列表
-
MemTensor/MemOS
Self-evolving memory OS for LLM & AI Agents: ultra-persistent memory, hybrid-retrieval, and cross-task skill reuse, with…
-
zhu1090093659/dsh-web-ui
Plugin and skin collection for DeepSeek Harness (DSH) Web UI - task board, git graph, right-side panel, remote mobile UI…