Browse agent skills

Data extraction tools for restaurant sites: public-facts research, photos (Instagram, TripAdvisor) and menu import from Cluvi, pictures or PDFs

Context usage breakdown for Pi — shows how the model context is split across system prompt, tools, MCP tools, skills, memory files, and messages.

Record pi agent sessions, play them back deterministically, and splice them into composable recipes

Claude Code skill: read, search and reply to your own Telegram chats, send files, and manage signed-in devices (mtcute)

Browse, search and install SkillHub skills inside DeepSeek Harness. / 在 DeepSeek Harness 内浏览、搜索并安装 SkillHub 技能。

Verify you are talking to LibertAI inference running in an attested confidential VM, then talk to it directly.

Localization for YAGE: message descriptors, an i18next-backed service, and localized text for renderer, UI, React, dialogue, and inventory

SLA-aware request scheduling for LLM serving, with a mock engine, an API server and benchmark tools

The attention engine's pure core: the contract, the lenses, the brief, the reconciliation plan, the prompt text, and the HTTP client. No database, no model, no React.

The `@dev.fast/whiteboard` package provides the `whiteboard` CLI for headless review authoring, sharing, and agent trace capture. It requires Node 24. An npm install also provides `review` as an alias for `whiteboard`. `whiteboard server` and `whiteboard

AEAP Composition Boundary Adapter for LangGraph. Thin wrapper that attaches the Runtime Decision Point to LangGraph node invocations.

Configurable usage and cost reporting for Pi coding agent

Typed Next.js SDK for Asqie CMS: fetch client, cache tags and webhook revalidation.

Motion graphics rendered from code with Claude; engine, checks, skill and an 8-lesson fact-checked course

Prepackaged DeepSeek Harness (dsh) runtime for core-darwin-x64

Durable customer approvals for AI workflows, plus Pushary notifications and operator questions.

n8n community node for the Omaha Homes For Sale CRM API (Real Estate Webmasters). Manage leads, groups, sources, agents, favorites, saved searches, notes, webhook subscriptions, and more.

Monocrawl for n8n: public social media, search and e-commerce data from 70+ platforms on one API key

n8n nodes for Synci: read bank accounts, balances, holdings and transactions, and trigger workflows on Synci webhook events.

n8n nodes for SmartChat: WhatsApp marketing with contacts, messages, newsletters, automations and SmartChat AI.

Neural microphone noise suppression for the browser: a Web Audio node backed by a streaming 48 kHz model

In-process model routing for LangChain: a chat model that picks one of several candidate chat models per request.

Vercel AI SDK provider for the Momento LLM gateway — one credential, every allowlisted route, with token vending and refresh handled for you.

An Agent Skill that turns complex topics and AI outputs into clear explanations, diagrams, interactive single-file HTML, and video storyboards. Inspired by Andrej Karpathy's ideas on understanding LLM outputs.

Policy enforcement for LLM agent tool calls, decided from each call's message ancestry

Local, read-only Codex and Claude Code API-equivalent usage dashboard

Fathom Charts client: real-time order-flow indicator streams (WebSocket) and history (REST), for Node and browsers.

n8n node for PostEverywhere: schedule and publish social media posts to Instagram, TikTok, YouTube, LinkedIn, Facebook, X, Threads, Pinterest, Bluesky, Telegram, Discord and WordPress.

DeepSeek Harness plugin: Claude Code-style prompt history, double-Esc picker, selection quote/copy & right-click paste | DeepSeek Harness 插件：Claude Code 风格上下键历史回溯、双击 Esc 历史搜索、选中文本引用/自动复制与鼠标右键直接粘贴

Paperclip adapter that runs OpenCode on Ollama Cloud models and retries issues right after a spent session or weekly limit resets.