Report by The Next New Thing.
Scroll for the report ↓
A collection of agent skills built for technical marketers and founders — conversion optimization, copywriting, SEO, analytics, and growth engineering. From Corey Haines (Swipe Files / Conversion Factory). Works with Claude Code, Codex, Cursor, Windsurf, and any agent on the Agent Skills spec.
View on GitHub →CRO, copywriting, SEO, analytics, and growth engineering, packaged for AI agents.
Aimed at technical marketers and founders, with a companion “Coding for Marketers” guide.
Works with Claude Code, Codex, Cursor, Windsurf — anything on the Agent Skills spec.
MIT-licensed; contributions welcome via PR.
If you do any marketing, your AI agent can run real marketing playbooks — landing-page CRO, SEO, copy — instead of generic advice.
Matt Pocock’s personal agent skills, straight from his .claude directory — small, composable, and model-agnostic. The pitch: where GSD, BMAD, and Spec-Kit take over your process, these stay small and easy to adapt so you keep control. At ~135k stars it struck a nerve.
The actual skills Matt Pocock uses daily for real engineering — not vibe coding.
Easy to adapt and combine, not a heavy framework that owns your process.
npx skills@latest add mattpocock/skills, then pick the skills you want.
Work with any model; based on decades of engineering experience.
You can drop a respected engineer’s actual day-to-day skills into your agent — a lightweight alternative to the heavier “process” frameworks.
A free and open-source video editor for web, desktop, and mobile, positioned as the open alternative to CapCut. It’s being rewritten from the ground up with a Rust core, a plugin-first architecture, and an MCP server for AI agents; the classic version runs today at opencut.app.
View on GitHub →Free, MIT-licensed video editing for web, desktop, and mobile.
A new Rust core targeting one codebase across browser, desktop, and mobile.
An Editor API, third-party plugins, and an MCP server so AI agents can drive it.
The classic version runs now at opencut.app while the rewrite lands.
You get a genuinely free video editor you can use in the browser or self-host — no watermarks, no subscription, and soon scriptable by AI agents.
Apple released an official, open-source CLI that creates and runs Linux containers using lightweight virtual machines, written in Swift and optimized for Apple silicon. Each container gets its own VM for fast start and strong isolation — a credible, native alternative to Docker Desktop on the Mac.
View on GitHub →Each container runs in its own lightweight VM — fast start and strong isolation.
Optimized for M-series chips on top of the Virtualization framework.
Pulls and runs the standard Linux container images you already use.
Open source under Apache 2.0 — a native macOS codebase, not a Linux port.
Familiar container run-style commands that fit existing dev habits.
If you build on a Mac, you can run Linux containers natively without Docker Desktop — faster startup, stronger isolation, and no licensing strings.
NVIDIA's open-source security scanner for AI agent skills. The research it cites shows 26.1% of skills contain vulnerabilities and 5.2% show likely malicious intent. It runs 64 vulnerability patterns across 16 categories — prompt injection, data exfiltration, supply chain — to answer one question: is this skill safe to install?
View on GitHub →Prompt injection, data exfiltration, privilege escalation, supply-chain, tool poisoning, and more.
Fast static scan plus optional LLM semantic evaluation.
Git repos, URLs, zip files, directories, or single files.
Queries OSV.dev for real-time vulnerability data with offline fallback.
Clear severity labels; outputs to terminal, JSON, Markdown, or SARIF.
Before you let an agent skill run on your machine, you can scan it and get a risk score — a basic safety check for the wild-west world of installable skills.
A developer ran skillspector scan against Angular’s official developer skill, got a 100/100 CRITICAL result, and filed the full report as an issue on the Angular tracker (Jun 15, 2026) — a concrete, real-world example of the scanner catching something in the wild.
Addy Osmani's set of production-grade engineering skills for AI coding agents. Seven slash commands map to the full development lifecycle — /spec, /plan, /build, /test, /review, /ship — encoding the quality gates senior engineers use so the agent follows them consistently across every phase.
View on GitHub →Spec → plan → build → test → review → ship, each activating the right skills.
/build autoGenerates the plan and implements every task in one approved, test-driven pass.
Designing an API triggers the API-design skill; building UI triggers the frontend skill.
/plugin marketplace add addyosmani/agent-skills.
Small atomic tasks, tests as proof, faster-is-safer shipping.
Drop these into Claude Code and your AI agent stops freelancing — it works the way a disciplined senior engineer would, from spec to ship.
I've used Zapier across multiple businesses for over 10 years. They've been doing automation reliably long before this moment — and the MCP service brings that same reliability to any AI agent you're building with.
Try Zapier MCP →Exposes 8,000+ tools as MCP resources any agent can call.
Claude Code, OpenCode, Cursor — anything that speaks MCP.
Enable the tools you want in Zapier, grab the MCP endpoint, paste it into your agent.
Zapier handles auth, rate limits, and reliability for you.
A PM skills marketplace: 68 PM skills and 42 chained workflows across 9 plugins, from discovery to strategy, execution, launch, and growth. Each skill encodes a proven framework and walks you through it step by step. Built for Claude Code and Cowork; compatible with other assistants.
View on GitHub →Discovery, assumption mapping, prioritization, strategy, launch, growth.
Grouped by job — discovery, strategy, execution, launch, growth, and more.
/discover, /strategy, /write-prd, /plan-launch, /north-star.
Teresa Torres, Marty Cagan, and Alberto Savoia methods encoded as steps.
Skills activate when relevant; force-load with /plugin-name:skill-name.
If you do product work, you get battle-tested PM frameworks running inside your AI assistant — structured decisions, not just faster docs.
A giant, community-curated collection of publicly available IPTV channels as M3U playlists, organized by country, language, and category. One of the most-starred repos on GitHub; the catalog is continuously checked and updated by contributors.
View on GitHub →Channels grouped by country, language, category, and region.
Works with VLC, Kodi, and most IPTV apps via a single URL.
CI continuously verifies which streams are still live.
Unlicense — use the data however you want.
6.7k forks; channels added and pruned by contributors daily.
You get a free, ready-made worldwide TV playlist you can paste into VLC or any IPTV app — no subscription, no account.
One CLI that lets your AI agent read and search Twitter, Reddit, YouTube, GitHub, Bilibili, and XiaoHongShu — the platforms that normally block scrapers or charge for API access. The pitch: all tools open source, all APIs free, install with a single line you paste to your agent.
View on GitHub →Twitter, Reddit, YouTube transcripts, GitHub, Bilibili, XiaoHongShu, RSS.
Open-source tools; the only optional cost is a ~$1/mo proxy for server use.
Paste an install URL to your agent and it sets itself up in minutes.
Cookies stay on your machine; nothing is uploaded.
Maintainers track platform changes and patch access methods as they break.
Your agent can finally summarize a YouTube video, read a Reddit thread, or check what X is saying — without you wiring up paid APIs for each site.
A context-compression layer for AI agents. It compresses tool outputs, logs, files, and RAG chunks before they hit the model — 60–95% fewer tokens with the same answers. Ships as a library, a proxy, and an MCP server, local-first and reversible.
View on GitHub →Six compression algorithms; local-first and reversible.
Drop-in library, transparent proxy, or MCP server.
Compatibility matrix for Claude Code, Cursor, OpenAI, LangChain, and more.
headroom-ai on both; install in seconds.
Designed to cut tokens without degrading model output quality.
You fit far more into a context window and pay for far fewer tokens — without changing your prompts or losing answer quality.
An AI-agent research skill that searches Reddit, X, YouTube, Hacker News, Polymarket, and the web in parallel, scores results by real engagement — upvotes, likes, money — and has an AI judge synthesize one grounded summary. Zero config to start; a wizard unlocks gated sources in 30 seconds.
View on GitHub →Reddit upvotes, X likes, YouTube transcripts, Polymarket odds.
Pulls many sources at once and ranks by real engagement signals.
One grounded brief instead of ten browser tabs.
Reddit, HN, Polymarket, GitHub work immediately; a wizard adds the rest.
Claude Code marketplace or npx skills add for 50+ hosts.
You get a recency-aware research brief built from what millions of real people are actually saying and doing — the live web, not stale editorial pages.
A free, open-source media library manager that connects your streaming services to a wide range of connected speakers. The server is the core — it runs on an always-on device like a Raspberry Pi or NAS and integrates tightly with Home Assistant.
View on GitHub →Unifies your streaming providers into a single browsable library.
Sonos, Chromecast, AirPlay, Snapcast, and more.
Runs on a Pi, NAS, or NUC — you own your setup.
Deep integration for whole-home audio automation.
Apache 2.0, with an active community of contributors.
If you self-host, this becomes the free hub that plays any of your streaming services on any speaker in the house — no vendor lock-in.
A regularly updated archive of extracted system prompts from the major AI products — Anthropic's Claude, OpenAI's ChatGPT and Codex, Google's Gemini, xAI's Grok, Cursor, Perplexity, and more. A goldmine for prompt engineers studying how production assistants are actually instructed.
View on GitHub →Anthropic, OpenAI, Google, and xAI, plus tools like Cursor and Perplexity.
New prompts added as products ship and leaks surface.
Published at asgeirtj.github.io/system_prompts_leaks.
CC0 — read, quote, and learn from them freely.
See how real assistants handle tools, refusals, and tone.
You can study exactly how shipping AI products instruct their models — a free masterclass in real-world prompt engineering whose patterns you can borrow.
New hot repos every week. Subscribe so you never miss the next drop.