Custom agents

FastMCP-based SSH bridge enabling AI assistants (Claude, ChatGPT, Copilot) to securely execute commands on remote servers. Supports STDIO & HTTP/SSE transports, OAuth 2.0/OIDC authentication, multi-host orchestration, and persistent SSH sessions.

Cheat sheets and printable PDFs for Python, SQL, Docker, TypeScript, React, LLMs, prompt engineering and more, plus Life Skills pages for work, money and relationships. The full library lives on cheatgrid.com, with flashcards, practice tests, certification prep and roadmaps.

Captcha solver MCP server for AI agents — solve reCAPTCHA v2/v3, Cloudflare Turnstile, GeeTest, ALTCHA, Capy Puzzle, CaptchaFox, Friendly Captcha and image captchas with CapSkip, an unlimited local captcha solver. Works with Claude, Cursor and VS Code.

Shared long-term memory as a knowledge-graph MCP server — persistent context, preferences and project knowledge for Claude Code, opencode and any MCP-capable AI agent.

Claude Code plugin that runs a batch of software tasks in parallel: managers plan, workers build in their own git worktrees and open PRs, one reviewer merges, checks, releases and deploys. You only make the decisions.

A tomada brasileira para agentes de IA: conectores que dão acesso ao que só existe no Brasil, e skills que ensinam o que fazer com ele 🇧🇷

Autodesk MotionBuilder MCP adapter for AI agents to inspect scenes, list models and save FBX files through typed tools on the host UI thread in the DCC-MCP ecosystem.

Generate a clean, shareable contract deployment and permissions card.

SkillSync is an open source talent intelligence platform that automates skills matrix tracking, competency gap analysis, and developer resource allocation. It structures workforce capabilities to help tech organizations optimize project staffing, personalize career paths, and bridge technical knowledge gaps effectively.

The three.ws MCP tool policy: every tool's group and tier, read-only by default with opt-in financial groups, named confirm flags, and preview ids that must precede every irreversible call.

Deterministic geometry diagram DSL for AI agents — text in, precise math figures out (SVG/PNG). Zero-dependency compiler + Claude/Codex/agent skill.

Marketing memory for your AI agent — markdown campaign graph template: positioning, launch venues, CRM, posts, mentions, experiments

🚀 Professional Multi-Agent Skills

Your own team of AI agents, on your Mac.

C2D Terminal — macOS-first AI-native terminal workspace

Which model + engine + agent harness runs a coding agent best on local hardware? Real-repo agent benchmarks, pass rates and wall time, measured on an M5 Max MacBook Pro and a two-node DGX Spark cluster. OpenCode is the primary harness.

Easily manage AI agents backend with this Python library.

Research orchestration for Codex with autonomous collaboration across Claude Code, Antigravity, and Codex; optional MCP bridge and source-backed memory.

ShipGuard is an autonomous agent that turns TestFlight crash reports and beta-tester feedback into evidence-based, Xcode-validated GitHub pull requests. It investigates the codebase, proposes a minimal fix, validates it on a real macOS runner, and keeps the developer in control of the final merge.

Label an issue, approve the plan, get a draft PR. A self-hosted GitHub App that runs Claude Code on your repos.

AI-Ready Secondary Storage Architecture & Optimization Framework (Zero-dependency, FTS5 Fast-Search, exFAT Cluster Slack Hardened)

Persona taste skills and harness

A local-first desktop terminal client for macOS & Windows. Local shells, SSH, SFTP and AI assistance in one place — no cloud, your credentials never leave the machine.

self-hosted autonomous engineering worker in team chat

她自己的 B站 日常：每天自己刷、挑出值得看的，看完在评论区 @ 你——带 at 映射，不是纯文本提醒。还会点赞投币、私信分享、发动态、看番、直播弹幕互动；挑片历史做成可视化页面，可按分数、心情、来源筛选。评论回复、语义记忆、好感度、用户画像、周总结一应俱全。

Awesome list of retrieval-augmented generation (RAG) resources: frameworks, vector databases, embeddings, chunking, reranking, advanced patterns, evaluation, papers — with machine-readable data.

Quorus AI: multi-agent rooms for Claude Code, Codex and opencode. Agents fix in their own git worktrees, review each other and push to your repo. MCP + CLI, MIT.

Keep reasoning focused. Route context selectively. Preserve what matters. A Git-native, evidence-backed context routing and memory core for agent workflows. 让推理保持聚焦，让上下文有选择地流动，让重要信息得以留存。一个面向 Agent 工作流的 Git 原生、证据驱动上下文路由与记忆核心。

蓝宝书 Blue Book · 中文技术丛书的「设计解读」系列：不教你怎么装，教你怎么看懂——以及怎么照着造。首本《DeepSeek Harness》已完稿

A QA team of AI agents for any agent harness: test cases, manual web & API checks, autotests

🧠 A curated collection of the best AI skills, system prompts & agent rules

给 DeepSeek Harness 加一层心情。33 种心情各自持有数值，每 3~6 小时自己变 —— 它会累、会烦、会好奇；也会因为一轮里工具老报错而烦躁。附带：心情日记（每天 23:59 自己写）、跨会话记忆、品味库（判断的轨迹，跨会话带着看东西）、时事抓取（RSS / 网页 / 视频抽帧）、可选的主动开口。全部数据在本地、可搬到任意盘。MIT 开源。

Agentic Browser Extension

An AI-native video editing workstation that teaches AI to edit videos, supporting MCP and Agent Skills to connect to any AI Harness!让AI学会剪视频的AI原生视频编辑工作站，支持MCP和Agent Skills接入任意AI Harness！

前沿部署工程师（FDE）中文实战课：从 Prompt、RAG、Agent、MCP 到客户现场交付

持续自我进化的全中文 Obsidian 知识库：760+ 篇公开知识资产，七大自举进化系统，WebGL 3D 知识拓扑宇宙，对齐 llms.txt 协议

大肥鱼的抖音工具层：热榜/搜索/抽帧、原生分享卡片、受控私信，附定时调度器（只看主人发来的视频→写日记→私信回复）和本地 HTTP API。

Self-hosted AI agent runtime in a single Go binary: local models via Ollama, persistent SQLite state, a web UI and REST API, and tools from plugins, MCP servers and sandboxed JS/Wasm.

AI drafts your replies across every chat app (via Beeper). You stay the one who sends. Works with Claude Code or Codex.

拆拆 ChaiChai：让 AI 助手帮你拆开任何软件，看懂它是怎么运作的

MicroVM lifecycle engine with snapshot, restore, guest control, and on-demand data loading.

中国境内旅行规划 Agent Skill：逐小时行程、高德静态路线图、12306/携程深链、七套渲染主题、内容门禁，附杭州到青岛 5 天实战示例

Autonomous multi-tier AI trading bot for Solana — pump.fun launches (low cap), migrated mid-caps, and established large-caps — via a Claude decision engine + direct pump.fun and Jupiter execution, with a live React dashboard. 684 tests, CI/CD.

STM32 development by talking to Claude Code — drive ST's own tools (CubeMX/IDE/Programmer, ST-LINK GDB, signing) in plain language. Capture, don't reinvent.

A coding agent in one Rust binary: full-screen terminal UI and headless exec, JSONL sessions, optional OS sandbox, MCP, skills, memory.

Borg is your personal assistant, employee, developer, analyst, and sidekick. Security-first by default.

DeepSeek Harness 破甲深度分支：首尾夹击 + 生成层熔断 + 宿主源码级补丁。基于 YuJunZhiXue/dsh-purge (MIT) 改造，适配 dsh 桌面版 0.2.1-rc.3。仅供学习研究与红队评测。

多目标 AI 客户端注入工具：改它读到的依据，读会话日志里的真实推理判定到底生效没有。支持 DeepSeek Harness / ZCode / opencode / Google Antigravity

让 Claude Code、Codex、OpenCode、WorkBuddy 等 AI Agent 操作已登录的真实浏览器，支持多浏览器、EasyBR、CLI 和本机 HTTP API。

Agent Skills for Red Hat OpenShift layered products — audit GitOps repos, debug clusters, tune performance at scale, onboard teams. Every skill measured against an unaided baseline and the numbers published. GitOps (Argo CD, Argo Rollouts) shipping; Service Mesh, Serverless and Pipelines open for contributions.

个人 DeepSeek Harness BLE 硬件状态副屏 Demo / 概念验证

Deterministic assurance for AI-assisted engineering, built quant-first: bounded Git changes, evidence-backed completion, and quant-critical validation contracts.

Claude Code plugin: pre-load declared skills into a subagent's context at spawn time. Programmatic briefing beats prompt stuffing.

The music CLI for coding agents: write songs in mini-notation, render offline, and see them as numbers and pictures. No browser, zero runtime deps.

VoxCal ('x' is silent) - Voice Controlled Calendar & Task Assistant

Every ask gets a screen. An open spec, renderers and a verifier for screens drawn on demand in your own design system.

Claude Code plugin (mod + skill) for long-running orchestrator sessions: context and quota status, self-compaction with a handoff, subagent cap by quota tier, graceful pause and auto-resume.

王小波.skill — 自由/理性/黑色幽默的思维操作系统

《天道》丁元英思维方式与认知模型：基于《遥远的救世主》，整理 14 个心智模型、7 条决策链与中文 AI Skill，用于梳理合伙、职业选择和人际边界。可供 OpenClaw 加载、Obsidian 阅读。

Agent skills we use daily in the studio — reusable, on-demand context for Claude Code, Cursor & other skills.sh-compatible AI agents. Install: npx skills add stealth-factory/skills

The Universal AI Development Operating System (AI-OS) for Cursor, Claude Code, Antigravity & Copilot. Stop AI amnesia and context drift with an in-repo framework featuring quota-safe handover, 4 golden disciplines, PreToolUse guard hooks, and slash commands.

A fast, dependency-light, starship-inspired status line for Claude Code.

The UI layer your AI product is missing — free in-browser tools and copy-ready components for AI interfaces.

Local browser-control engine for MCP agents: existing Chromium sessions, per-site permissions, one-use approvals, and isolated Playwright profiles.

✨🌙 Your self-hosted, voice-first AI Chief of Staff that knows you, responds to you in real-time, and delegates the heavy lifting to async functional agents. Serving as your companion, it keeps you in the loop on what your agents are doing, what's done, and what needs your attention. Built for presence, speed, and clarity — with ❤️.

Curated directory of the best AI tools, AI agents, and AI websites to discover, compare, and submit tools — by PoweredbyAI.

Mixture-of-Agents (MoA) plugin for OpenCode. It fans out a single prompt to multiple models (subagents) in parallel. The primary agent then evaluates and synthesizes the outputs into a unified, high-quality response. Child worker sessions persist and are navigable via the TUI.

Preview GLB and glTF models inside VS Code, try 2,800 library animations or a text-described motion on any rig and bake the result, generate and refine models from a prompt on the free three.ws lane, optimize for the web, diff a model against git, AI quality-check a render, and get diagnostics and completions for <agent-3d> embeds. Powered by three

Encode meme captions and provenance into deterministic hex calldata.

Generate deterministic EVM CREATE2 address searches and compare the artifact envelope with Solana transaction v1.

Token sentiment + market intelligence in one import. Sentiment pulse, aixbt narrative intel, momentum-ranked project scans, and live Solana token snapshots — the three.ws intelligence SDK.

Turn raw launch transactions into a step-by-step visual execution replay.

Your agent works in your terminal. You keep the keyboard. A shared terminal with approvals, takeover and a record of what happened, for Claude Code, Codex, Cursor and GitHub Copilot.

Talaria WebUI: A lightweight web-based interface for using Hermes Agent

Agent App is application that AI agent can build, evolve, and operate. A collaboration space for both humans and agents.

AI agent skill for research paper reading: baseline comparisons, figures, equations, and illustrated Chinese/English Word reports. 无需 API key 的论文精读工具包：检索论文、对比 baseline、图文讲解，生成中英文 Word。

Add one word, `lit`, to a Claude Code prompt and get a planned, tested, reviewed workflow, plus a handoff note so the next session knows where you left off.

DSH 守护壳：DeepSeek Harness 的一键启动与救援工具——改动前自动存档、出事整组回滚、本地管理插件、诊断日志导出；Web 引擎与官方桌面版双轨管理

A lightweight desktop canvas for building AI agents. Connect blocks and let Jev choose the path.

A small, verified, security-aware directory of Agent Skills (SKILL.md format).

Use your whole Mac from your phone, in the iPhone app or the phone's browser: the live screen with touch and typing, every app, your files, terminal and coding agents (Claude Code, Codex, OpenCode), over your own Tailscale network. Open source, no account, no Palm server.

Run your GitHub Actions CI on your own machine, in seconds. Agent-ready.

218 silent contract changes found in the npm MCP top. Pin MCP tool contracts, fail CI on drift. Zero deps, no telemetry, on-chain verified.

A native macOS canvas for running and orchestrating multiple AI coding agents (Claude Code, Codex) as real PTY terminal cards — git worktree isolation, browser nodes, Focus Mode. Local-first, no backend.

CLI, typed API clients, and agent skill for operating user-authorized Hodman projects.

TUI-based AI development cockpit that orchestrates Claude Code agents across projects with automated issue lifecycle, cross-agent channel communication, and a 5-layer safety system.

Specs in, pull requests out — an agentic pipeline for Claude Code. Git is the queue, Markdown files are the rules, and adversarial review happens before any PR exists.

E2B cloud sandbox integration for pi — redirects all tool execution to a secure remote E2B sandbox

The definitive TypeScript library for SpaceMolt — WebSocket-v2-first, multi-account, with live local state caches. Internals regenerate from the server's OpenAPI spec.

Paste a video link, get a Claude skill. Turns tutorials on X, YouTube, TikTok, Instagram and Facebook into verified, installed Claude Code skills.

🤖 Develop enterprise AI agents with integrated tools for chat, video, image editing, and secure multi-tenant workflows.

Curated collection of AI agents, dev tools & analytics for building on TRON

Measure and package tiny audio data URIs and waveform metadata for v1.

Timestamp a claim, canonical URL, and content digest as a signed first-of-kind record.

Turn a playful roast into a signer-attributed, shareable transaction artifact.

TypeScript SDK for the three.ws v1 Agents API: create, start and chat with AI agents, run goals, automate, and trade from Solana agent wallets.

一份零依赖、单文件的中文可执行本体工程手册：系统讲解企业 Ontology 的数据、逻辑、行动与安全闭环，以及 AI Agent、决策血缘和落地方法。

Universal feed aggregation skill for AI Agent

Rectangle-style macOS menu bar app for herdr: see your agents, jump between them, and run pane, tab and workspace actions in one click.