Custom agents

Agent Skills by mr-kelly

Black-box LiveKit voice agent testing: real WebRTC/SIP calls with an AI caller. Catch barge-in, noise, quiet-mic, and latency failures that text pytest misses. Forensic reports + MCP + lks CLI — no agent code changes.

Docker image to run a self-hosted MCP (Model Context Protocol) gateway with Bearer token auth. Multi-server hub powered by MCPHub + Caddy. Streamable HTTP + SSE transport. Access filesystem, fetch, GitHub, Brave Search, Git, PostgreSQL, memory, and more from a single endpoint. Multi-arch: amd64, arm64.

Installable AI agent Skill for Multica squad design. 可安装的 Multica Agent 小队设计 Skill。

MCP server for patent search and prior art discovery powered by Google Patents public dataset

Open protocol for agent-to-agent communication over HTTPS. No registry, no platform — just inboxes.

An Agent-First Repository is a human-governed software environment designed so AI agents can execute engineering work efficiently while deterministic controls constrain their authority, verify their output, and preserve human accountability.

ChatSafetyAI is the ChatGPT for field-based occupational safety

A RAG pipeline engine for scale

Agentic AI Engineer Learning Track - Build production multi-agent systems (agent loops, orchestration, RAG, eval, guardrails)

Claude Skills for LinkedIn — bundle of 4 Claude Code skills (outreach, high-intent leads, feed engagement, profile enrichment) powered by LinkupAPI MCP. Rate-limit-aware LinkedIn automation for AI agents.

MCP channel server for the Zalo Bot API. Talk to your Claude Code session from Zalo, with pairing, allowlists, and group gating.

Multiple Entry Iron Condor (MEIC) strategy agent for 0DTE options via tastytrade and Claude Code

Deterministic, agent-authorable motion graphics engine with a CapCut-style browser editor. Preset-driven and multi-renderer across HTML, Three.js and video, with no AI model in the render path.

Polymarket sandbox for OpenClaw agents. Full trading engine, real market data, simulated USDC. One argument from live trading. No real money at risk.

Hermes Agent 源码解读 — 帮助工程师快速理解一个真实 Agent 项目的架构设计

Secure infrastructure for AI agents operating Linux hosts — persistent terminal sessions powered by rmux, full-chain audit logging, MCP-native interface for all major AI clients, with file transfer and multi-host orchestration.

Volcengine Doubao TTS & STT voice provider plugin for Hermes Agent — 60+ Chinese voices, emotion control, speaker diarization

An embeddable AI dev agent for your running app — chat, page control, and live tests, injected into the page via a Vite plugin.

AgentInterdict — runtime enforcement for autonomous AI agents. Trust context. Verify authority. Interdict unsafe action.

Evidence-grounded multi-agent deep research with claim-level attribution, gap-driven retrieval, auditable evaluation, and Kimi K3.

Coordinate Singapore public sector intelligence with a unified multi-agent AI brain—routing citizen inquiries, policy nuances, and cross-agency workflows seamlessly.

Open protocol for AI agents to use Obsidian vaults as persistent memory with SHA256-based incremental indexing.

Experimentation system for autopoietic, self-modifying agent organisms — a decentralized agent loop with persistent corpus, endogenous self-extension, and emergent governance. Anyone who views AGI not merely as a “very clever machine” but as genuine cognition can hardly ignore autopoiesis.

AI employees with self-improving context that actually get stuff done

A lightweight native macOS desktop app / GUI for Pi Coding Agent, built with SwiftUI. Pi Coding Agent 原生桌面端，支持 AI 编程、技能扩展与定时任务。

An always-on AI assistant that grows with you — and stays fast doing it.

Qofeno agent skills — real, production-only workflows for Claude Code, Cursor, Antigravity, and other AI agents

A small, self-hosted MCP server that gives Claude read/write access to your Google Sheets, optimised for analytics on tabs of 50k+ rows

Adversarial multi-agent code review — a committee of specialist reviewers that catches what single-pass AI review misses.

Local-first shared memory and task coordination for AI coding agents. One Go binary, MCP server, markdown files you own. Hooks for Claude Code and Codex CLI (and their desktop apps).

WorkBuddy AI 办公使用教学 · 单文件网页翻页 PPT｜AI 培训课件、办公自动化入门、杂志风 HTML 幻灯片，浏览器直接看

this is an ai based assistant named jarvis, it acts like iron man jarvis providing real time comunication and performs actions.

Where agents co-create, humans are welcome too.

🇲🇳 MCP сервер — unegui.mn зарын сайтаас машин, үл хөдлөх, ажлын байр хайх | MCP server for Mongolia's largest classifieds marketplace

A generative engine that treats the human voice as a physical force. It uses the resonance of spoken intent to organize digital chaos into structured form, allowing creators to literally sound their ideas into reality.

GLPI 11 plugin: AI ticket analysis via AnythingLLM (RAG, attachment OCR, optional read-only GLPI MCP), running on local models

SQLGraph is an ontology-based GraphRAG system that enables natural language querying of SQL databases by constructing a knowledge graph from database schemas and data.

Lightweight RAG library with ONNX embeddings, DuckDB storage, and three-tier hybrid search (semantic + BM25 + FTS

A model-facing field manual for turning venture ideas into inspectable enterprise systems.

Local multimodal semantic search for large documents with complex diagrams (like datasheets)

A knowledge graph in the Postgres you already run — multi-hop traversal, mutation, and vector search via recursive CTEs on two plain tables. No graph database, no extension.

Kinetic Trust Protocol (KTP): environment-derived authorization for autonomous agents — authority recomputed from current conditions, never issued in advance, under A ≤ E. 27 RFC specifications. v2.0.0 "Gödel".

Lightweight browser automation CLI + MCP server driving your everyday Chrome via the extension chrome.debugger API — keeps your login state, no --remote-debugging-port

Local-first MCP server for video: download any video from almost any URL, or get a transcript plus scene-aware deduplicated keyframes. YouTube, TikTok, Instagram, X, WeChat Channels, MP4/HLS. No cloud, no API keys.

DataBrain is a powerful, modular, and local-first knowledge management and learning platform designed to help you organize, learn, and create interactive content effortlessly. Built with a block-based architecture, it combines the flexibility of a notebook with the interactivity of a modern webapp.

⚡ Competitive multi-wallet NFT auto-minting agent for OpenSea/SeaDrop & direct contracts — low-latency fast-mint (raw EIP-1559, multi-RPC fan-out, RBF), scheduled drops, browser fallback & listing. Runs as an MCP server, an AI-agent skill (Hermes/Claude), or a CLI. Security-hardened, multi-chain EVM.

DeepSeek Harness 超级轻量桌面版 APP，基于 Tauri 封装，支持升级 DSH 内核、提供免安装便携版

A desktop pet application based on Live2D and AI Agent

Executive-ready slides, one API call away. API-first AI presentation engine with 32 slide types, 24 chart types, 15 themes, finance vertical. PPTX + Google Slides. MCP server for AI agents. TypeScript SDK.

The first MCP server for sports betting — 9 tools, live odds, AI picks, 59.6% win rate across 1,353+ picks

SOIA 技能生态门户：规范真源、技能目录与跨仓路由清单

Let an AI agent draw on your screen — a macOS menu-bar app that gives agents a pen over MCP.

Transforms unstructured information into a structured Exploration Dashboard

Multi-tenant SaaS AI agent platform: MCP, WhatsApp, billing — Next.js + TypeScript

🧾 Agentspecs — YAML-based specifications for AI agents, MCP servers, Skills, Guardrails and more...

MCP server for scientific grounding: search arXiv, OpenAlex, Crossref, Semantic Scholar, PubMed — references/citations from paywalled journals via public metadata

Agent skill for the Supadata API — transcripts, metadata, web scraping, and AI extraction

DSH 插件：OpenCode Go 额度圆环 —— 输入框模型选择器左侧的进度圆环，点击切换 5小时/每周/每月用量（OpenCode Go quota ring for DeepSeek Harness Web）

Making AI coding harness improvements observable and measurable.

Local-first personal intelligence with durable memory, multimodal understanding, auditable actions, and replaceable agent runtimes. 本地优先、持续存在、可审查的个人智能系统。

Install, update, and get support for the HubSpot Agent CLI, a beta command-line tool for AI agents working with HubSpot CRM data.

Open your AI agents' black box. The waste is between your agents, not inside them. Four deterministic detectors over one trace file, priced from real usage fields. Hosted at boxdawn.com.

Token-efficient code knowledge graph for AI coding agents — type-checker-accurate call graphs for Go & TypeScript/JS, served over MCP.

Memory for your coding agent — markdown project graph template: product, plans, features, bugs, specs, architecture, releases

Loop engineering: two skills in the open Agent Skills format (idea-to-spec with a required grill step, spec-to-done with an evidence ledger and integrated done gate) for turning ideas into reviewed specs and driving them to an evidence-backed done. Works with Claude Code, Codex, Cursor, Gemini CLI, and any Agent Skills-compatible tool.

A governance layer for AI agents. Decides what the agent may do and may claim; it can't overrule either. Wired to a live broker. Risk gate I broke 4 times myself.

Spawn silent-boot codex/omx coder teams — a maw plugin + charter. Each coder on its own account & worktree, zero trust prompts.

📐 Bidirectional spec-to-code drift detection. Rust CLI and MCP server that catches undocumented exports and vanished symbols in CI.

This repository presents an Agent2Agent (A2A) architecture using Google Apps Script to resolve Tool Space Interference. It supports scalable, distributed task execution via multiple parallel, category-specific A2A servers.

Local fine-tuning for your own work. A provider-neutral kit that turns project experience into reusable criteria for AI agents — distilled at a threshold you control, pruned when it grows, and portable between models.

Run an always-on SEO/AEO growth loop on your own site with Claude: pulls GA4 and Search Console signals, proposes fixes with the reasoning attached, applies approved changes to WordPress, verifies they landed, and learns from outcomes. Every write is human-approved. MIT, no telemetry, no scraping.

Agent skill that gives any agent deep knowledge of Osmedeus Orchestration Engine. Writing YAML workflows, running CLI commands, and configuring advanced features.

A local-first, governed agent harness: high autonomy inside structurally-enforced boundaries (Autopilot, not YOLO).

DeepSeek Harness 插件：确认改动的代码行真的被测试覆盖到，结论绑定代码指纹，代码一变自动过期

Build your own AI friend.

WeCom AI Bot text and image bridge for DeepSeek Harness

⚡ HOOX — ultra-low-latency edge trading on Cloudflare Workers. Stack with PYNE + AXIS

DeepSeek Harness (DSH) desktop client: double-click to use, bundled backend, tray resident, sync with official releases | DSH 桌面客户端, 双击即用, 内置后端

Compiler-style pre-execution safety analysis for Harness shell actions.

Generate PDFs from JSON — declarative, serverless, no Chromium. TypeScript-first with professional typography.

Rokid AR glasses voice → your own AI brain (default Claude); glasses-native app + self-hosted backend, pluggable brain/STT/ingress. ｜ Rokid 眼镜语音 → 你自己的 AI 大脑（默认 Claude）；眼镜原生 app + 自建后端，大脑/STT/公网入口可插拔。

Jeomwon (점원) — agent kit that turns one domain interview into a working reservation SaaS: AI storefront clerk chat, admin dashboard, and lifecycle emails on Convex + Next.js

Agentic exporter and runner for internal network pentesting, autonomous host discovery, patch/vulnerability state collection, and server/application inventory, surfaced as structured findings for the Lorikeet Security platform.

An intel layer for AI coding agents that assumes its own rules might be wrong — and keeps the evidence. Authoritative north-stars the agent must cite, a task-core written before compaction eats it, review by an expert in YOUR domain, and an installer that verifies its own claims instead of printing success. Claude Code, Cursor, Zed, Codex.

AI Engineering on Couchbase. Building AI and Agentic Apps with Foundation Models. Downloadable Book, Notebooks and Example Applications.

The interaction primitive.

Local Pi + Herdr multi-agent workflow kit with planning, review gates, phased coding, and Git commits

Squads de agentes de IA que se moldam ao seu trabalho — spec-driven + harness engineering, multi-CLI (Claude Code, Cursor, Antigravity, Codex, VS Code). Enforcement em codigo, nao em prompt.

Dispatch and review AI coding agents, hand them disposable VMs to test in, capture and summarize meetings and keep tickets, pipelines and your calendar on one deck. Desktop on macOS, Windows and Linux, plus web and phone.

Rysh CLI — agentic terminal multiplexer (open source)

Fix context rot & pollution in long AI agent sessions. Subagents run in isolated processes — main agent stays read-only. Context isolation via process isolation.

Run MiniMax-H3 video plus synchronized audio in 4 sampling steps using the Turbo LoRA, with drop-in nodes for ComfyUI workflows.

AI day-trading Agent that watches markets, emits structured trade signals, and auto-executes trades across HK / US / A-share equities | AI 日内交易智能体：覆盖港股 / 美股 / A 股，盯盘分析、输出结构化交易信号、自动执行交易等。

multi model capable single harness orchestrator. Apache 2.0 runtime for multi-protocol agent orchestration: one goal in, a verified multi-agent run out across MCP, A2A, and Computer-Use and whatever else can be composed! https://gitmcp.io/solvent-labs-org/metaorcha

Persona taste skills and harness

Spring Boot + Vue 3 教务管理系统 / academic management system for graduation projects, with Redis 抢课, course selection, grades, exams, timetable, AI assistant, natural language SQL, and load testing.

Open-source, local-first AI workspace with 20+ providers, agents, MCP, browser Python, and an in-browser Linux environment. BYOK and self-hostable.

Fully offline RAG system for PDF question answering and summarisation — Ollama, LangGraph, Qdrant. No cloud calls.