Curated

Open-source plugins that connect Claude Code, Codex, opencode, DeepSeek Harness, QwenPaw, QwenWork and WorkBuddy to one MemoryLake. Local auto-memory syncs up in the background and is recallable from any machine or client. Configure once, shared by every harness. MIT.

Official Rust CLI for MemoryLake: upload files, store facts, search memory across projects, manage agents and talk to them over A2A. Every command prints JSON for jq and scripts. MIT-licensed, one-line install on macOS, Linux and Windows.

Jev is a classifier, not a chatbot: feed it an input plus an output schema and get a probability per option in ~200ms. Ryan Vogel demos email triage (1,700 emails for 18 cents), lead scoring, support routing, video clipping and browser control, plus where it fails.

Billy Howell runs a 6,000-reader newsletter on a Grok Bot agent team: chief-of-staff bot, sales agent, overnight routines — plus a 4-week starter plan and the best business models to copy.

Matt Pocock's daily-driver agent skills for Claude Code and Codex: /grill-me to align on intent, /tdd red-green-refactor, /code-review, /diagnosing-bugs. MIT-licensed, one command to install.

Jack Roberts fixes Claude's AI-slop design in 3 levels: a 2,000+ design-system gallery, a critique-agent design loop for pixel-perfect HTML email, and a local design OS with style recipes.

Y Combinator's QM gives every employee an isolated agent sandbox, memory, and keychain — plus shared Slack channels. Swap Pi, Codex, or Claude Code. MIT.

Greg Isenberg's agent-first playbook: sell work as a service priced like labor — pick an already-paid workflow, ship a minimal agent, pilot it, productize in 30 days.

Cody Schneider maps a real Facebook-ads agent on Greg Isenberg's podcast: Airbyte + ClickHouse data stack, AI-generated creative, auto publish/kill/promote — plus a concrete startup idea to build it around.

YC Partner Diana Hu on the AI-native company: AI as your operating system, a fully queryable org, the 1,000x engineer, and why middle management disappears.

Palantir launches Agent Engine + Agent SDK at DevCon 6: context items, events, and effects turn agent loops into multiplayer distributed state machines.

Palantir designers share design patterns for legible, trustworthy human-agent collaboration: presenting agent reasoning without overload, with clear decision attribution.

Claire Vo breaks down the four agent loop types (heartbeat, cron, hook, goal), then live-builds a daily PR reviewer in Claude Code and a self-validating skills loop in Codex.

HeyGen's open-source video framework (35k stars, Apache-2.0): deterministic MP4s from HTML/CSS and seekable animations. Ships 20 Agent Skills — product launch videos, website-to-video, PR explainers, beat-synced music videos — so one prompt lets Claude Code take a video from plan to final render.

A 28k-star GitHub template: run /clone-website <url> and your AI agent screenshots the site, extracts design tokens, writes component specs down to exact computed CSS, then dispatches parallel builders in git worktrees to reconstruct it all — as a clean Next.js 16 + shadcn codebase.

A design-guidance system for AI coding agents by jQuery UI creator Paul Bakaus (46.7k stars): one skill, 23 commands (polish/critique/bolder/distill…), live browser iteration, plus 46 deterministic detector rules that run with no LLM — calling out the exact AI tells: Inter everywhere, purple-blue gradients, cards nested in cards.

The exact Claude Code setup YC President Garry Tan uses daily, at 122k stars: 23 specialist roles plus 8 power tools that turn AI into a full engineering team — a CEO who rethinks the product, an eng manager who locks architecture, a designer who catches AI slop, a QA lead driving a real browser, a security officer running OWASP audits, and a release engineer who ships the PR. Free, MIT.

A 63k+ star collection of Agent Skills: 13 single-purpose skills spanning taste injection (layout, type, motion), image-to-code, redesign audits, and brand-board generation — one npx skills add away in Claude Code, Codex, or Cursor, so your AI stops shipping boilerplate-looking UI.

300+ DESIGN.md design-system analyses extracted from real websites — Airbnb, Apple, Stripe, BMW and more. Pick one, drop it into your project, and your AI coding agent inherits the colors, type, and spacing. You don't need to know any of it.

A design-system service delivered as an MCP server: your AI agent picks a design system extracted from real products — 80+ production-ready components each, semantic tokens, UX principles, and WCAG checklists — so generated UI stops looking like generic AI output.

A community registry of hand-crafted React components and templates: thousands of polished components across 50+ categories, with CLI and MCP integration for AI coding tools — so your AI starts from real, beautiful UI instead of the bland AI average.

A library of 500+ website templates and animated components as ready-to-use prompts across 20+ categories: copy, paste into Lovable, Bolt, Cursor or Claude, and ship an animated landing page in minutes. Free tier available; premium designs via subscription.

A free, MIT-licensed terminal AI SEO audit tool for Claude Code: 25 sub-skills, 18 parallel subagents, 50+ /seo commands — outputs a 0-100 health score with a prioritized action plan, replacing Ahrefs/Semrush-style subscriptions.

A Claude Code skill that makes AI follow production-grade Next.js App Router i18n patterns — translation dictionaries, clean default-locale URLs, hreflang, localized JSON-LD, and multi-language sitemaps.

An open-source (Apache-2.0) desktop design workspace that turns your own coding agent — Claude Code, Codex, Cursor, Gemini and 21+ others — into a design engine for prototypes, landing pages, dashboards, slides and HTML video. Local-first, BYOK, no vendor lock-in.

Greg Isenberg and Elie Steinbock turn lean-startup and Toyota-style feedback loops into copy-ready workflows for SEO, Facebook ads, and product feedback — with a live Search Console SEO-loop demo.