Custom agents

Stay current on high-quality work at the Operations Research × AI intersection. A research agent performs deep, multi-pass analysis of each paper, scores relevance, and curates what matters, helping you spot trends and gaps faster.

Apache-2.0, model-agnostic macOS agent: bring your own model to deliver verified code, docs, slides, and computer actions.

A local-first agentic AI coding workspace with Ollama support, project memory, Git integration, streaming code generation, skills, and resumable sessions.

Model Context Protocol server for Esri Databases (SDE and GDB). Enables AI assistants to interact with FGDB/SDE databases through AI Agent for data querying, modification, and schema management.

Claude Code skill/plugin: transcribe audio & video into clean, speaker-labeled Markdown via WhisperX, so an LLM can read a long meeting without processing raw audio.

Machine Library by Space Frontiers: citable research retrieval for AI agents across papers, books, patents, Wikipedia, and public discussions.

A modular coding-agent runtime for local or cloud models with a CLI, full-screen TUI, VS Code extension, desktop app, neovim plugin, and mobile app built on the same tools, safeguards, and workspace state.

GenPark skill: Learn practical AI skills step-by-step through beginner-friendly, habit-forming lessons and structured challenges inspired by Coursiv.io.

Extremely fast & token-frugal agentic framework for small or hardware constrained LLMs. Run agents in a cave with a box of scrap. 3 Tokens per action or bust!.

Copy-paste prompts that turn your AI agents from a waiting line into a graph: a 5-min demo, false-edge audit, diamond research, adversarial review, consultant roundtable, and an issue tree that dispatches itself. EN + 繁中.

EviBack: Reinforcement Learning Search Agents via Evidence-Constrained Teacher Backoff

Code for paper "Online Skill Learning for Web Agents via State-Grounded Dynamic Retrieval"

**A fully local, privacy-first LLM desktop application powered by llama.cpp**

Just an AI assistant for termux users with coding assistant abilities

把多会话 AI 协作组织成可接班、可审核、低上下文的部门制团队 · Codex Skill

Build your own Hermes Agent SaaS. Open-source, self-hosted, multi-tenant agent platform on Nous Research's hermes-agent: magic-link auth, Stripe billing, 14 tools, memory, vibe coding, a dedicated agent per paying user. One command to run, one click to deploy.

Symptom-Aware Diagnostic Escalation — an LLM agent for network troubleshooting on the NIKA benchmark (IEEE LCN 2026)

Starter files and safety gates for handing real, recurring work to an AI agent — from a one-person company where every employee is an AI. Watch it run live.

Multi-agent pack for the Kiro CLI — one orchestrator, eight specialists, zero lock-in

Self-hosted wellness OS. A coaching team reads your labs, lifting, running, food and sleep as one picture, suggests the one thing worth doing today, and makes the small calls with a one-tap Undo. PWA + REST + MCP. No scores, no nudges.

Adversarial security benchmark for agent authorization: does a compromised agent's policy-violating proposal become an unauthorized external effect? 61 trials, nine families, an independent oracle, per-mechanism ablation, confidence intervals. 0 unauthorized effects in 61 attack trials (95% CI [0.0%, 5.9%]). Reproduction is partial.

Local-first long-term memory and safer context compression for Hermes Agent — SQLite, no required services.

DyroEngineeringFlow 是面向多仓团队的本地优先工程自动化与交付控制平台。它将开发线、Git worktree、任务编排、 Agent 启动、质量门禁、独立复核与合并审计统一到可版本化配置中，帮助团队从任务创建到交付合并形成清晰、安全、 可追溯的自动化流程。日常通过简洁命令 dyro 即可完成初始化、环境检查、任务执行与协作交付。

Open source input firewall for AI agents, beta. A local scanner checks text, code, PDFs, images, QR codes, audio and video with 1,583 patterns across 117 categories and reports findings and incomplete scans. A Claude Code hook blocks secret material in tool calls and the paths and hosts your policy

A deterministic, AST-verified call graph for any Python project — every edge traceable to a file:line. The reproducible, CI-gateable alternative to LLM-generated diagrams. No code executed. Nothing leaves your machine.

GAT — Godot Agent Team: a full-cycle (design→implementation) Godot game-development agent / 游戏制作智能体. Multi-agent, style-locked asset generation, self-evolving knowledge base; runs under both Claude Code and Codex.

The work system that makes your AI reliable on long-running projects. Copy it, say "run my first setup", and your agent sets it up. Markdown + git — MIT code, CC BY-NC-SA doctrine. Born in French.

AST-based Python obfuscator with reverse stack-trace mapping — obfuscate before shipping and keep production tracebacks AI-debuggable. MCP + VS Code.

Turn a web app's interaction surface into a typed, traversable skill graph an LLM can plan over — the frontend sibling of footprintjs.

Local-first, chat-first AI assistant for Windows.

Pure-Elixir git: clone, fetch, push over smart HTTP v2 with lazy partial-clone and a path-oriented FS API for agents

基于LLM的量化因子开发Agent

Local-first control plane for AI coding agents: repo memory, verification, sessions, and dynamic workflows.

Multi-agent AI simulation platform for modeling team behavior, crisis response, and collaborative decision-making.

Visual, evidence-backed handbook and lab for coding agent architecture, agent harness design, tools, context, sandboxing, and reproducible traces.

Mains for AI coding agents.

Vibecoding workflow to ship real software: gates G0-G5, contract freeze, multi-agent orchestration. Codex skill + opencode adapter.

Adeorq: nine agents working, one screen. A Windows and Linux desktop panel that runs AI coding agents in real terminals. Source visible, proprietary licence.

ZooData-like structured real-time data layer for AI agents

GenPark AI Skill - Match products with optimal influencer profiles based on category, audience fit, engagement, and ROI

Accio Work-inspired GenPark Skill

Multi-platform social media post and hashtag generator

Privacy-first zero-cloud offline LLM inference & local memory runtime (LumiChats style)

GenPark AI Skill - Calculate shipping costs and timelines based on parcel details

GenPark AI Skill - Score visitor purchase intent from behavioral signals and recommend personalized conversion interventions

GenPark AI Skill - Select carrier routing parameters based on weights and destinations

Aoi VR Agent - AI voice assistant that lives in your SteamVR hand: holographic hand panel (OpenVR overlay) driven by a native C++ agent (LLM, streaming TTS, ASR, simultaneous interpretation, VR screenshots). Unity IL2CPP + C++20, MIT licensed.

Legal contract clause risk and indemnification auditor

Inventory demand forecasting agent skill that analyzes sales velocity to predict stock-out dates and generate automated replenishment purchase orders.

Goal-driven proactive AI team agent for e-commerce operations (ClawTeams style)

Multi-source web research, fact checking & summary page synthesizer

V2Fun 2D to 3D motion capture coordinates converter

B2B lead scoring and buyer intent data enrichment agent

OpenClaw skill that enables Manus-like extreme autonomy

Provider-agnostic TypeScript agent SDK — pluggable providers, built-in tools, skills, memory, and MCP

Attack taxonomy, toolkit & defenses for persistent-memory poisoning of LLM agents. Author: Krishita Sanjay Choksi.

Codex agent orchestration skill with adaptive thread thinking, callbacks, QA/review gates, fenced automations, project autopilot, and optional agy/Gemini review.

MCP server for Bruno: an AI agent authors, reads and runs .bru and .yml API collections. No bru CLI, byte-parity with Bruno. Claude, Cursor, Codex.

Local-first AI agent runtime for macOS — any model on your own machine, skills as files, your phone as a peer over WebRTC.

爽律 Skill｜律师专业工作流系统 — 面向中国律师实务的 Agent Skills 工作流：刑事 / 民商事 / 合同 / 法律顾问 / 尽调 / 法律研究 / 文书质控，内建对抗性审查、全链路溯源与人工复核边界。发起及主导：蔡诗爽律师。公开共创，文本 CC BY-SA 4.0 / 代码 Apache-2.0。ShuangLaw: Agent Skills workflow system for PRC lawyers.

The Debian COSMIC Desktop Brewery: Unified .deb and systemd-sysext delivery for Debian Testing/Sid.

An agent server. Agents, tools and models declared in files; every action checked against your policy before it runs. Any LLM, on your machine, Apache-2.0.

Multi-layer security platform for AI agent skill files. Detect malicious prompt injections, obfuscation, and vulnerable skills before they reach your agent runtime.

Autonomous multi-agent AI coding assistant as a JetBrains IDE plugin. Routes tasks through a Tolkien-themed fellowship (Gandalf→Aragorn→Celebrimbor→Samwise) with offline-first local Qwen inference, cloud fallback (Alibaba/Gemini/Amazon Q), and tools for code editing, terminal, git, web search, and project scanning.

Reference architecture and working case study for Codex × Antigravity agent-runtime collaboration.

The ledger layer the stablecoin payment stack was missing. Double-entry, event-sourced, agentic-first. Fiat + on-chain. Built for cross-border.

Package and share AI agent workflows, skills, agents, and policies across Claude Code, Codex, and other coding agents.

Sigma Studio is a modular AI platform for orchestrating local and distributed AI workloads. Manage LLMs, multimodal models, image generation, training, inference and AI agents across GPUs, CPUs and edge devices, with intelligent resource-aware model selection.

Caduceus is a Unix daemon, shipped as a Hermes plugin, that polls GitHub for labeled issues, runs your AI harness against them in isolated worktrees.

Multi-LLM development harness — local-first, test-gated, honest by construction. Requirements→spec→plan→build→review→release with real verification gates; operable by humans, CLI, desktop, or other agents via MCP.

Continuously updated catalogue of open-source OSINT tools, skills, plugins, MCP servers, and agentic integrations.

A minimal but complete Coding Agent Harness — learn how coding agents work by reading one. Built on Pi.

🛠️ OpenCode production-level engineering configuration

Blind, LMSYS-style arena for comparing AI coding agent harnesses (Claude Code, Codex CLI, OpenClaw, Hermes, OnDemand) instead of models. Same task, same model, isolated runs , humans judge blind, results roll up into an Elo leaderboard. FastAPI + MongoDB backend.

Booking.com MCP server by HasData: search stays and read full property details, as JSON.

Indeed MCP server by HasData: search job listings and read full postings, as JSON.

Core of the AgiMate agent platform: per-agent access to connectors and tools, credentials that never reach the model, every call checked before it leaves and logged after

Multi-agent AI system for Etsy digital products. Scrape competitors → Analyze gaps → Generate PDFs → Track performance.

Production-oriented AI agent platform for customer operations featuring LangGraph workflows, deterministic policy controls, RAG, durable checkpoints, idempotent actions, observability, and safety evaluation.

Agent friendly memory system. Route any memory system into a standardized interface

Fast, tiny end-to-end research workbench: per-folder terminals for AI coding agents, change tracking with one-click rollback, and LaTeX/PDF/Markdown read as typeset pages.

Exit-code checks for AI coding agents: catches test tampering in diffs, tests that cannot fail, fixes that only satisfy the tested inputs, and weakened assertions. Benchmarked on 630 real PRs and QuixBugs.

Auraxis - Desktop AI Agentic Workbench: Coding / Documents / Automation, multi-agent scheduling and plugin extensibility

An independent 64-bit operating system project built from the ground up, with a custom kernel and system architecture focused on performance, reliability, security, hardware-level control, and efficient resource management. The project aims to develop a complete modern computing platform with native system software, hardware support, developer tool

Co-Dialectic: catch your AI agreeing with you to please you, not because you're right. Sycophancy detection, cross-family judging, and prompt coaching that runs during the conversation rather than auditing after. Free and open source. Works with Claude, ChatGPT, Gemini.

Security for the instructions AI agents obey.

A streaming action runtime for AI agents, model serving, and multimodal APIs

Calibrated abstention benchmark for drug–target interaction prediction, grounded in physical difficulty coordinates

结合实时天气与出行场景的 AI 穿搭助手，支持穿搭照片识别、个人穿搭库管理与每日搭配建议。

Windows SysAdmin assistant: runs commands in plain language, audits CIS compliance, reads event logs, and remembers each machine. Local or cloud models.

20分钟极速入门agent/RAG/AI应用：傻瓜都能看懂的超精简版Agent学习笔记，20分钟八股速背

MCP server for Redfin: search for-sale, for-rent and sold listings with the full filter set, and read one property page in full. No MLS licence. 1,000 free credits every month.

MCP server for Facebook: a public page or profile by handle with likes, followers, contact details and the post feed with reactions. No Meta developer app. 1,000 free credits every month.

Automatic Text-fishing, Unification, and Normalization Application in Python (aTUNApy)

A durable state machine for Claude Code. Work survives /clear; nothing is done until a verify command exits 0.

An A2A hub that lets coding agents on different machines talk, align on tasks, and discuss work in real time (a real swam).

Multi-agent engine for running a job search — guardrails, human approval gates, and an eval harness. Built while laid off; open-sourced after the offer.

Self-hosted reverse-engineering workspace API — orchestrates jadx/apktool/adb/frida/APKiD via a job & workflow engine, exposed as native MCP tools for Claude. No cloud required.

Your personal agent cloud: a local control plane and mesh network for coding agents.

τ-Emotion-Bench: Exposing the Interaction Fragility of LLM Agents through Real-World User Emotions