29 October 2026

JetBrains agentic IDE built around the pi agent

An agentic IDE that wraps the pi coding agent in a full interface. Work is isolated into separate workspaces, and the agent keeps project knowledge through living specs and skills.

• Each workspace is a git worktree with its own branch and folder, so several agents can work in parallel without touching your checkout.
• It has Monaco editor tabs, a file tree with live git change marks, terminals, and review comments that can be sent to a chat as context.
• The top bar shows where the branch stands against origin and its pull request, including CI checks and the review decision.

A collection of design skills for AI coding tools

A collection of standalone skill folders for Claude Code, Cursor, Codex, Windsurf, and other tools that read markdown rule files. For now there is only one skill: landing-page-design.

• landing-page-design builds a landing page from scratch: initial questions, page structure, conversion copy, and a visual system (typefaces, spacing, radius, motion).
• For Claude Code, the skill folder is copied to .claude/skills. The README also gives steps for Cursor, Codex, and Windsurf.
• To upgrade an existing site, the README points to a separate repo called redesign-skill.

A complete software development method for coding agents

A skills framework that makes a coding agent ask first, write a spec, and draft an implementation plan before writing code. Skills trigger automatically, so no special commands are needed.

• The spec is shown in short chunks that are easy to read and approve before the implementation plan is made.
• The plan stresses red/green TDD, YAGNI, and DRY, then runs through a subagent-driven-development process that checks the result of each task.
• The README says the agent can often work on its own for a couple of hours without deviating from the plan. Install guides cover Claude Code, Codex, Cursor, Gemini CLI, and more.

A personal AI tutor for lifelong learning

An AI tutoring platform built with Python and Next.js. Release v1.7.0 (28 Oct 2026) brings modular installation, a refreshed learning workspace, block notebooks, and narrated video lessons.

• Earlier releases mention reviewed learning-path changes, grouped assessments, bounded tutoring memory, and question-bank practice.
• It includes document reading with Office previews, source figures sent to vision models, and interfaces in several languages such as German, French, and Ukrainian.

A team workspace: email, chat, docs, tasks, and agents

A single interface that unifies email, messages, docs, tasks, agents, calls, and CRM with shared AI memory. Everything is @linked and searchable.

• It is built from modular blocks: multi-account email, channels, Linear-style tasks, CRDT-based collaborative docs, and a 2D canvas.
• All blocks share one backend, and references between docs, tasks, messages, and emails are stored as a bidirectional graph.
• It was designed in NYC and Toronto and used by the team of about 15 for two years.

A design skill that avoids the AI-generated look

A design skill for Claude Code, Codex, Cursor, and OpenCode. It deals three themes from a catalog of 50, builds the page around what the visitor came to do, and checks it in a browser.

• The result is linted in a real browser against tells of machine-made pages, such as an eyebrow above every heading. A fresh-context reviewer has the last word.
• Rules come in three tiers: the Floor (contrast, focus, phone layout) never ships broken, Reflex gates may be overruled with a reason, and Finish checks never block a ship.
• /hallmark redesign rebuilds a page and keeps its content, while /hallmark audit only reports.

Design vocabulary for coding agents: polish, critique, audit

Design commands for coding agents: polish and bolder to steer, critique and audit to review. Strongest on interfaces that already exist.

• critique scores a design against usability heuristics, while audit checks accessibility, performance, theming, and responsive behavior with findings ranked P0 to P3.
• The slop detector uses 62 deterministic rules, with no LLM and no API key.
• /impeccable init runs once per project and writes PRODUCT.md, which later commands read.

An open-source AI agent that works across your devices

An open-source AI agent that connects your computers and phone and uses the models and subscriptions you already have. Your computer becomes the server, with no server rental.

• A session can be continued from another device, and model providers on another computer can be used. Remote desktop on the phone can operate a computer.
• Claude Code, Codex, and Pi are brought together in one interface, and the session stays in Cindy when the harness is switched.
• Project files stay on their own computer. The README says this is not whole-disk sync.

A desktop workspace for persistent AI agents

A local-first desktop app where each agent has its own workspace and conversation, with queues, file transfers, an embedded browser, and agent-to-agent messaging. It is a development preview.

• It supports Codex App Server and Claude Code, plus Grok CLI, OpenCode, Gemini, Cursor CLI, and Cline CLI through ACP.
• FIFO message queues can be paused, resumed, and cancelled, and survive crashes. An embedded browser can be controlled by agents.
• The README warns that agents run with danger-full-access and approvalPolicy: never, so run only agents and tasks you trust.

A collection of system prompts for many AI models

A repo of 32 prompts for 26+ models: system prompts, personas, workspace policies, AGENTS.md files, and Ollama Modelfiles. The README states research and educational purposes.

• Grouped by model family: free models on OpenCode Zen, open-weight models (GLM, DeepSeek, Kimi, Qwen), and closed models.
• The README forbids use to harm others or to violate platform terms.

A userscript that fetches download links from cloud drives

A JavaScript script that gets file download addresses from cloud drive services such as Baidu Netdisk, Aliyun Drive, and Quark. It is modified from the 网盘直链下载助手 script with more features.

• The README lists nine supported services, with a cleaned-up UI and support for several downloaders.
• It can push links to IDM and ABDM and copy Aria2 or cURL commands. Test version 1.1.4 adds support for 光鸭云盘.
• Links come from each provider's public API. The README states the project does not lift service speed limits.

Official Expo team skills for AI agents

Official AI agent skills from the Expo team for building, deploying, upgrading, and debugging Expo apps. Expo docs, Expo CLI, and EAS CLI remain the source of truth.

• Claude Code and Codex install it as a plugin through their official marketplaces. Cursor, OpenCode, and other agents use the skills CLI.
• Example requests: an Expo Router screen with tabs and modals, an EAS workflow that builds previews on pull requests, and an upgrade to the latest Expo SDK.
• Agents choose the skill from the task context and each skill's description.

WiFi sensing for presence and vital signs

A WiFi sensing platform that reads signal disturbances (CSI) from low-cost ESP32 sensors to detect people, breathing, heart rate, and movement without cameras.

• It recognizes presence and occupancy, activities such as walking and falls, sleep quality, and room changes through RF fingerprinting.
• It connects to Home Assistant over MQTT with the --mqtt flag, or to Apple Home, Google Home, Alexa, and SmartThings as a Matter Bridge.
• Each node ships 21 entities: 11 raw signals and 10 inferred states, such as someone-sleeping and bed-exit.

A self-teaching book with Python in Jupyter notebooks

The book 自学是门手艺 by 李笑来 uses programming, mainly Python, as the entry point to teaching yourself. Chapters are Jupyter notebooks.

• The table of contents covers why self-teaching matters, values and operators, control flow, functions, strings, containers, files, and the official Python tutorial.
• Part two covers deliberate practice and why to start from writing functions.
• A Markdown version has existed since 23 March 2019, and a guide explains installing Jupyterlab to read the book.