Building blocks and infrastructure for developers

SWE-Gym provides 2,400 real software-engineering tasks drawn from 11 Python repositories, together with repository context, executable environments, and test-based verification. Researchers can use it to generate trajectories and train agents or solution verifiers. Reproducing the experiments also involves the published dataset, prebuilt Docker images, and instructions for OpenHands or MoatlessTools.

This template combines Tauri 2, the Next.js App Router, Tailwind CSS, and TypeScript with preconfigured Biome, Rust formatting checks, and GitHub Actions. Developers can run the frontend inside a Tauri window and package a statically exported build for release. Because a packaged Tauri app does not include a continuously running Node.js server, the frontend must use static site generation rather than Next.js server-side rendering. The author also notes that Next.js may be more framework than many Tauri apps need.

agent-skills-eval tests an Agent Skill against the same prompts in two modes: with the skill loaded and without it as a baseline. A judge model grades both outputs against defined assertions, producing JSON and JSONL artifacts plus a static HTML report that shows measurable lift, failures, timing, and tool-call behavior. The project includes a command-line runner and a TypeScript SDK, supports OpenAI-compatible chat backends and custom providers, and can run in CI. Users supply both the target model and the judge model.

MDA is a Markdown-based specification and CLI for maintaining agent-facing documents from one source. Authors write a single .mda file and compile it into SKILL.md, AGENTS.md, MCP-SERVER.md, and CLAUDE.md outputs, with JSON Schema validation, typed relationships, dependency metadata, content digests, and signature fields. It is intended for authoring and CI checks rather than as an application runtime dependency. A bundled signature verifier, dependency resolver, and central registry are not yet included.

Create general-purpose agents in TypeScript with streamed events, filesystem or shell tools, persistent multi-turn sessions, and middleware for compaction, retry, and turn tracking. The SDK also covers subagents, MCP servers, on-demand skills, virtual filesystems, and React or Vue chat integration on top of Vercel's AI SDK.

This repository turns invoice-style Excel data into Neo4j nodes and relationships with pandas, then shows how to convert graph relationships into matrices. It preserves a legacy py2neo v3 and Neo4j 3.x teaching path while adding a modern example with the official driver, vector indexes, and GraphRAG-style retrieval. Current Python and Neo4j environments may require changes to the pinned legacy code.

Inspect Evals gathers benchmarks for coding, cybersecurity, safeguards, reasoning, knowledge, and other model capabilities. Researchers can run individual tasks with Inspect AI, improve included evaluations, or register externally maintained implementations. The project recommends Python 3.11 or 3.12 for the most reliable development and execution experience.

Skald packages document ingestion, chunking, embeddings, vector storage, retrieval, and cited chat behind a configurable API. Teams can tune search, reranking, models, and query rewriting, then evaluate the resulting RAG setup through the same platform and access it from several language SDKs. It can run as a hosted service or through Docker in your own environment; a fully local deployment also requires self-hosted LLM inference and embeddings.

Spring AI Agent Utils gives Java developers a reusable toolkit for building coding assistants and multi-agent workflows with Spring AI, covering file and shell access, code and web tools, task state, memory, and extensible skills. It is distributed as Maven modules with examples, requires Java 17 or newer, and includes an A2A module for delegating to remote subagents.

color-expert is an installable skill for Claude Code, Codex, Cursor, Copilot, OpenCode, and other agents. A concise SKILL.md provides frequently needed facts, corrections, and practical guidance, while an index routes the agent to 144 deeper references on color spaces, APCA and WCAG accessibility, palettes, pigment mixing, historical theory, tools, papers, and transcripts. Trigger evaluations, realistic task prompts, and maintenance guidance help keep the collection organized. It is intended for explanations, color reviews, and implementation advice.

Pyloid lets Python developers package web frontends and Python backends as desktop applications for Windows, macOS, and Linux. It includes a project generator and common desktop facilities such as windows, tray integration, storage, timers, servers, and build tooling. It is a framework for teams that want to keep their existing web technology choices while delivering installable desktop software.

This benchmarking toolkit runs ARC-AGI tasks through configurable adapters for providers such as OpenAI, Anthropic, and Gemini, with rate limiting, retries, saved submissions, and scoring built in. It supports both single-task debugging and asynchronous batch runs. Full evaluations require separately cloned ARC-AGI data, and hosted models require their respective API credentials.

Velox brings a Tauri-style desktop application model to Swift, connecting HTML or modern web frontends to Swift backends through typed IPC. Its CLI can initialize projects, run watched development sessions with optional frontend HMR, build releases, and create macOS application bundles; simple static interfaces do not require Node.js. The implementation still compiles a Rust FFI layer around Tao and Wry, and the newer Swift-first runtime remains incomplete, so the event-loop APIs are currently the primary path.

Add-to-Homescreen is a drop-in JavaScript library that shows platform-specific instructions for installing a website or PWA on iOS, Android, Windows, or macOS home screens. It supports more than twenty languages and can redirect users from several in-app browsers to the system browser. The site must already meet basic PWA requirements, including an app icon and web manifest.

Use openapi-to-skills to turn a local or remote OpenAPI specification into Markdown documentation organized for on-demand agent reading. It groups resources, operations, and schemas, with filters and custom templates for shaping the output. Specifications with external references need to be bundled before conversion.

afm runs Apple's on-device Foundation Model or Hugging Face MLX models on a Mac and exposes them through an OpenAI-compatible server, a web chat interface, single prompts, or shell pipelines. Its gateway mode can also discover and proxy local backends such as Ollama, LM Studio, and Jan, with separate OCR and embedding endpoints available. The main platform requirements are macOS 26 or later, Apple Silicon, and enabled Apple Intelligence.

This progressive sample shows how to protect an ASP.NET Core Web API with the Microsoft identity platform and call it from a WPF desktop application after user sign-in. Later chapters add on-behalf-of access to Microsoft Graph, personal Microsoft accounts, and proof-of-possession tokens. Running the examples requires .NET and Visual Studio on Windows, a Microsoft Entra tenant, and an eligible user account, so it is intended as learning material rather than a finished application.

simple-modern-uv is a compact template for Python 3.11–3.14 projects, paired with an Agent Skill that can create a new package, modernize an existing repository, or apply later template updates. It sets up uv, Ruff, BasedPyright, and GitHub Actions, and can migrate projects from Poetry, setuptools or pip, and PDM while tailoring license and PyPI publishing files.