Shape agent capabilities into your workflow.

Video Podcast Maker turns a topic into a narrated video workflow: it researches the topic, drafts the script, tracks assets, produces voice-over and subtitles, mixes audio, and renders with Remotion for platforms such as YouTube, Bilibili, Xiaohongshu, Douyin, and WeChat Channels. Free sources can resolve automatically; paid image or video generation requires a quote and explicit approval, and the script still needs repeated human polishing.

This library gives medical and biomedical researchers more than 550 skills spanning evidence discovery, protocol design, clinical statistics, bioinformatics, academic writing, and lab operations. The skills work with several SKILL.md-compatible agents, including Claude Code and Codex, and the project states that each one is evaluated through its medical research skill audit framework before release.

AIlice is a general-purpose autonomous agent that runs from the terminal or a web interface. It can read and analyze documents, conduct focused research, write and execute scripts, perform system-administration tasks, and support voice interaction. Cloud models impose few local hardware demands, but the project says effective local use generally needs a model of at least 70B parameters and two RTX 4090 GPUs, with Ubuntu offering the best compatibility.

Future AGI is a self-hostable open-source platform for observing and improving LLM or agent applications. It combines tracing, evaluations, simulations, datasets, guardrails, an AI gateway, and prompt optimization in one feedback loop. Traces use OpenTelemetry, while the gateway exposes OpenAI-compatible HTTP and integrates with many providers and frameworks. A full deployment includes PostgreSQL, ClickHouse, Redis, RabbitMQ, and Temporal, so it best fits teams that need unified quality and observability infrastructure.

Awesome LLM Skills is a curated directory of reusable LLM and agent skills across development, Notion workflows, creative work, media, and other tasks. It also collects setup guidance and resources for Claude Code, Codex, Gemini CLI, OpenCode, Qwen Code, and related platforms. Some entries live in the repository, while others link to external skill projects.

claude-blog gives solo publishers and content teams a repeatable blog workflow: start with strategy and a content brief, then draft, check SEO and facts, add structured data and internal links, and pass reviewer gates before delivery. It also supports multilingual publishing, topic clusters, and optimization for AI citations.

This repository packages portable Codex skills for two practical workflows: auditing a fast-moving project before commit or deployment, and applying stricter reliability habits during coding tasks. Clone it, copy or symlink the desired skill folders into your agent’s skills directory, then invoke them by name. The current catalog centers on the wtf audit skill and the better-codex reliability overlay.

Capture clicks, typing, navigation, and waits with a visual recorder, edit and replay the workflow, then export it as reusable MCP commands or Skills files for AI agents. BrowserWing also offers an agent-friendly CLI, an HTTP automation API, session and cookie management, and built-in scripts that return structured data. It requires an accessible Chrome or Chromium installation.

effective-html is a collection of Agent Skills for producing self-contained visual HTML deliverables. The general html skill handles broad pages, html-diagram creates full-screen architecture and system diagrams with SVG-first presentation, and html-plan builds structured planning pages. Each skill bundles a local example corpus to guide the intended style. Users can install the full collection or one skill, then optionally annotate the rendered result with Plannotator or share it through a separate service.

Install NoPUA to steer coding agents through repeated failures with perspective changes, evidence checks, new approaches, and follow-up inspection instead of threats. It encourages agents to state uncertainty honestly, verify completion, and prepare a responsible handoff when a problem remains unresolved. The skill supports several coding assistants and seven languages, although its full rule set may overlap with an existing workflow stack.

Loom is an experimental Rust coding-agent research project with a REPL, file and code tools, server-side model proxying, conversation persistence, and remote execution components. Its repository explicitly restricts use to Geoffrey Huntley and warns that the software is unstable, unsupported, and subject to breaking API changes.

Happier keeps Claude Code, Codex, OpenCode, and other coding-agent sessions running on a computer while making them accessible from a phone, browser, or desktop client. Users can follow, take over, fork, replay, or hand off live sessions without discarding context. Communication is end-to-end encrypted, and the relay can use the hosted service or a self-managed deployment. Remote access requires the target machine and its daemon to remain online; it complements existing agent CLIs rather than replacing their model services.

FIM One is a self-hosted agent platform for connecting APIs, databases, and MCP servers behind one execution layer. Teams can create connectors from OpenAPI specifications, a conversational builder, or MCP, then run multi-step work through a ReAct loop or dynamically generated DAGs. The same agent core can be delivered as a standalone assistant, embedded copilot, or central hub. Its connector strategy explicitly spans global SaaS and China-focused systems such as Feishu, WeCom, DingTalk, and domestic databases, and deployment requires a compatible LLM endpoint and credentials.

LangChain4j-AIDeepin is a self-hostable web platform for teams building intelligent business assistants. It combines multi-character chat, image generation, RAG knowledge bases, visual workflows, MCP tools, speech input and output, and conversation memory, with separate admin and user interfaces plus REST APIs. Docker deployment is documented, while supported capabilities vary across model providers.

App Store Preflight checks an iOS or macOS project before review by inspecting the Xcode project, source code, configuration, and store metadata. It identifies the app type, loads a matching checklist, scans for known rejection patterns, and reports severity, affected files, and remediation steps. Its references cover common cases such as subscriptions, user-generated content, kids, health, games, macOS, and AI apps. Metadata inspection uses the asc CLI and assumes its canonical JSON layout, so fastlane projects may need path adaptation.

Inkeep Agents lets non-technical teams design agents on a drag-and-drop canvas while engineers define the same agents in a TypeScript SDK, with two-way synchronization between both formats. It supports customer assistants, internal copilots, and multi-agent workflow automation, along with MCP credential handling, embeddable chat components, tracing, and Vercel or Docker deployment. The platform can be self-hosted, but it uses Elastic License 2.0 with supplemental terms rather than a permissive open-source license.

BrowserGym exposes open-ended browser tasks and benchmarks such as MiniWoB, WebArena, WorkArena, and VisualWebArena through a consistent Gym interface. Researchers can install only the core environment or add benchmark-specific packages, then run agent–environment loops with Chromium controlled through Playwright. It is designed as research infrastructure rather than a consumer automation product, and individual benchmark suites require their own setup.

Agentshire is an OpenClaw and QClaw plugin that maps agents and sub-agents into visible 3D citizens. Its browser town reflects task dispatch, collaboration, streamed dialogue, coding activity, and completion animations, while users can chat through the same interface. A map editor and character workshop let people customize the environment and NPCs. Starting the OpenClaw Gateway opens the local town after installation. Compatibility is version-sensitive: OpenClaw 2026.3.13 is recommended, while 2026.4.x is not yet supported.