6 November 2026

Graph context layer so AI agents read data correctly

A Python library from NVIDIA that builds a context layer from enterprise metadata, reconciles definitions across systems, and exposes it to agents through MCP, CLI, and API.

• Generates ontologies from metadata and connects existing ontologies or semantic models.
• Gives agents business meaning for metrics, entities, and relationships, and reconciles inconsistent definitions.

Small, composable agent skills from real engineering work

A set of agent skills the author uses every day, kept small, easy to adapt, composable, and usable with any model.

• Installs as a plugin for Claude Code, Codex, or GitHub Copilot, or with npx skills@latest add mattpocock/skills.
• After install, /setup-matt-pocock-skills runs once per repo to choose the issue tracker and ticket labels.

One Claude Code setup that works as an engineering team

A set of 23 Markdown slash-command tools that turn Claude Code into a virtual team: CEO, designer, eng manager, reviewer, QA, and release engineer.

• Quick flow: /office-hours writes a design doc, /review finds bugs that pass CI, and /qa opens a real browser to reproduce bugs.
• /autoplan reviews a plan end to end, then /ship runs tests, review, and the PR.

Xianyu listing monitor with Playwright and AI analysis

A multi-task monitor for the Xianyu (Goofish) marketplace built on Playwright and AI, with a web UI to manage tasks, accounts, and results.

• Each task has its own keywords, price, filters, and AI prompt, and can run on a Cron schedule.
• Notifies through ntfy.sh, WeCom, Bark, Telegram, or Webhook; main data is stored in SQLite.

Web scraping framework from one request to a full crawl

A Python framework that combines fast HTTP requests, stealth browsers, parsing, and concurrent crawling in one library, with an MCP server for AI agents.

• Supports JavaScript rendering, Cloudflare Turnstile handling, persistent sessions, and proxy rotation.
• Parses with CSS, XPath, text, and regex; crawl results export to JSON, JSONL, CSV, or XML.

Computer use CLI for AI agents via the accessibility tree

A CLI and C library that reads a desktop app's accessibility tree as JSON with a ref per element, so an agent can operate apps without screenshots.

• The agent clicks or types on a ref, reads the UI again to check the result, and repeats; there is no model inside the tool.
• Runs on macOS, Windows, and Linux.

Git worktree CLI as easy as branches, for parallel agents

A CLI for managing Git worktrees, designed for running many AI agents in parallel; worktrees are addressed by branch name with paths from a template.

• Three core commands: wt switch feat switches, wt switch -c -x claude feat creates and starts Claude, and wt remove cleans up.
• Includes hooks for create and merge, LLM commit messages, and a merge flow with squash and rebase.

Agent skills for Google products and Google Cloud

A repository of Agent Skills for Google products and technologies, including Google Cloud, from onboarding and IAM to multi-product solution skills.

• Installs with npx skills add google/skills, then you pick skills, target agents, and project or global scope.
• Includes solution skills such as a solution-architecture workflow, RAG with GKE and AlloyDB, and deploying AI agents on Google Cloud.

Browser CLI for AI agents with accessibility tree snapshots

A native Rust CLI that gives an agent a browser driven one command at a time; snapshot returns the page as compact text with a ref such as @e1.

• The agent acts on a ref, reads the page again to check the result, and repeats; the browser stays open between commands.
• Installs through npm or Homebrew, or runs without install with npx agent-browser open example.com.

Agent skills for images, video, audio, and 3D via Scenario MCP

A set of Agent Skills that teach coding agents to create content through the Scenario MCP server, plus expert skill teams for Blender, Maya, ZBrush, Unreal, and Unity.

• Skills pick the right model, price before spending, and keep characters and brands consistent.
• Expert skill teams drive DCC software and game engines installed on the user's machine.

Agent Skills collection for web design, video, and retrieval

A collection of production-ready Agent Skills for Claude Code, Cursor, Codex, and other coding agents, covering web design, knowledge retrieval, image generation, and more.

• Example: motion-showreel renders short videos from code and delivers an MP4 plus editable source.
• Other skills: web-design-engineer, gpt-image-2, beautiful-article, kb-retriever, and web-video-presentation.

Open-source voice AI model family from Microsoft

A Microsoft project releasing open voice models, including VibeVoice-ASR for transcribing long audio and VibeVoice-Realtime-0.5B with a range of voices.

• VibeVoice-ASR handles up to 60 minutes of audio in one pass and returns a structured transcript: who (speaker), when (timestamps), and what (content).
• Supports over 50 languages, user-customized context, vLLM, and a streaming version with hotwords.

Free local AWS emulator with no account or token

An open-source local AWS emulator for development, testing, and CI that provides AWS-shaped services on your machine without an account, auth token, or paid feature gates.

• Point the AWS SDK, CLI, Terraform, CDK, OpenTofu, or your test suite at http://localhost:4566.
• Presented as a drop-in replacement for LocalStack: swap the image and keep going.

Tool to create and burn images visible on a disc surface

An app that converts an image into data fitting the physical limits of CD, DVD, and single-layer BD-R, so a pattern shows on the disc surface after burning.

• Flow: create a project, design the surface with layers, preview, then export ISO, RAW, or WAV+CUE, or burn directly.
• Includes calibration from a disc photo and experimental live burning; use only blank discs you can afford to lose.