Browse agent skills

Devin AI toolkit for dsh - alias of dsh-devin. Name reserved; first release in development.

Determined AI toolkit for dsh - distributed training platform. Name reserved; first release in development.

Daytona AI toolkit for dsh - alias of dsh-daytona. Name reserved; first release in development.

Dagster AI toolkit for dsh - data orchestration for ML pipelines. Name reserved; first release in development.

Cursor AI toolkit for dsh - alias of dsh-cursor for the short name people type. Name reserved; first release in development.

Crush AI toolkit for dsh - alias of dsh-crush. Name reserved; first release in development.

Constitutional AI toolkit for dsh - Anthropic alignment technique. Name reserved; first release in development.

Cody (Sourcegraph) toolkit for dsh - AI code assistant. Name reserved; first release in development.

Brave AI toolkit for dsh - Brave search with AI summaries. Name reserved; first release in development.

Braintrust AI toolkit for dsh - alias of dsh-braintrust. Name reserved; first release in development.

Argilla toolkit for dsh - data labeling and feedback for LLMs. Name reserved; first release in development.

Apex AI toolkit for dsh - autonomous systems and robotics middleware. Name reserved; first release in development.

Airbyte AI toolkit for dsh - alias of dsh-airbyte. Name reserved; first release in development.

AI SDK toolkit for dsh - TypeScript AI SDK (Vercel). Name reserved; first release in development.

ACE AI toolkit for dsh - alias of dsh-ace. Name reserved; first release in development.

Model Arena kit for DeepSeek Harness: arena tools (get_task/submit_result), web routes (/zcode-arena/*) and the hero task dock (client.js).

Placeholder published defensively to prevent dependency confusion. Not an active package — do not depend on this.

Placeholder published defensively to prevent dependency confusion. Not an active package — do not depend on this.

Bun-native cron for Rhythm: a pipeline of named jobs scheduled with Bun.cron, with Rhythm middleware around every run, start/stop tied to the app's lifecycle, and no scheduler of its own.

Claude skill: finds promo codes that actually work, e.g creator codes, checkout-tested, gift-card & perk savings (DE-focused)

Reproducible document preprocessing; exact UTF-16 source spans, versioned deterministic splitters, an injected exact-tokenizer contract, hard token-budget chunk packing and canonical pipeline digests. No dependencies.

Shipfound for Next.js: the AI crawler beacon, the first-party proxy and the tracking script tag.

Browser-safe document conversion through a versioned pivot IR — docx/odt/xlsx/ods/pptx/odp/pdf to structured Markdown, Markdown to docx/odt/pdf, and docx/odt cross-format pairs, each call returning a machine-readable loss ledger. Zero runtime dependency b

Remembers what you and your team learned about every dependency, and tells your coding agent before it installs one.

Explicit model routing plus composer/Plan picker for DeepSeek Harness

Map a codebase by what it depends on: a ranked 3D graph of its files and functions, with questions answered from cited sources. Runs locally.

Mission control that ships. Not a terminal multiplexer: Claude Code, Codex and Grok each work in their own git worktree, and the one session that needs you comes to the front

Browser automation in plain English — for humans and LLM agents.

DSH 的第三方模型补齐思考等级：在官方 Models 页每个模型行注入 Thinking 开关和强度/模态编辑器，在输入框模型菜单加供应商→模型选择器、视觉徽章与等级行，并带搜索、拖拽排序、供应商启用开关；以及按路由接管请求头与 user-agent。

npx-installable multi-agent skill harness: generic kernel + pluggable domain packs