Absent Author is an open-source research-integrity toolkit with an evidence catalog, two agent skills, and supporting scripts for examining whether a research paper has been responsibly authored, steered, and checked by a human. The `paper-author-pass` skill guides authors through questions, claim–evidence mapping, reference and number verification, prose review, and a final checkpoint; `paper-slop-screen` gives reviewers writing, research, and quality grades, locations and quotations for findings, a quadrant classification, a review paragraph, and a note for an area chair. Its standalone tools lint surface patterns, verify references, compare quantities across a paper, and inspect PDFs for hidden or remapped text; the tools report candidates for human review rather than AI-detection verdicts. The repository also maintains a sourced evidence list with false-positive notes and paper-type exemptions. The skills can be installed for Claude Code, Codex, or another agent that reads `SKILL.md`, and the project is MIT-licensed.
An open-source Python tool for extracting local chat histories from AI coding assistants into normalized JSONL. It supports ten sources, including Claude Code, Codex CLI, Cursor, Windsurf, Trae, Continue, Gemini CLI, OpenCode, Cline/Roo Code, and Aider, and preserves messages together with fields such as code context, diffs, tool calls, timestamps, session IDs, project paths, and model names when the source stores them. The extractor detects common macOS, Linux, and Windows data locations, reads JSONL, JSON, SQLite, and Markdown transcript formats, normalizes each conversation into a shared schema, and writes one JSON conversation per line. SQLite databases are opened read-only; Cursor has explicit handling for several known storage shapes, while Windsurf and Trae use best-effort heuristics for undocumented schemas. It provides an interactive CLI as well as options for extracting all or selected sources, listing detected installations, adding search paths, and merging output into one file. The project uses only Python's standard library and requires Python 3.9 or newer. It includes guidance to scan extracted data for secrets and review proprietary code, API keys, and personal paths before sharing or using it for training. It is licensed under the MIT License.
AI Homepage is a Chrome/Chromium new-tab extension that uses Claude to generate a homepage from the user's browsing history. It ranks visited domains, opens up to eight selected pages in the signed-in browser, extracts headlines and links, and writes a React component that the extension renders as the new-tab page. Users can provide plain-English instructions for the page's content and visual layout, exclude hosts, and rebuild manually or on a 6-, 12-, 24-, or 48-hour interval. The extension runs a Claude Managed Agents session from its background worker. Its custom getHistory and getPageHtml tools access Chrome history and tabs, while the agent uses grep, read, and write operations in a sandbox to process page files and produce the component. It requires Chrome or another Chromium browser, Node.js 20 or later, and the user's Anthropic API key; the repository states that each build is billed to that key. The project is early and experimental, distributed as source code to build and load unpacked, and licensed under MIT. The README says that browsing-history summaries and cleaned HTML from up to eight pages are sent to the Anthropic API, that agent sessions and uploaded pages are deleted after each build, and that the stored API key can be read by anyone with access to the browser profile.
ai-iso-skill is a Claude skill that generates interactive isometric figures as single HTML files. It renders plain SVG without libraries: projection code maps box faces to SVG matrices, while paint order handles depth and a state object with render() drives interaction; CSS supplies motion. Before drawing, the skill has Claude determine what pressing a part should produce, such as typed text, a number, or a note. The resulting figures use a hairline drawing on a numbered plate and can include functional controls such as timer buttons or a light switch. It can be installed as a Claude Code plugin, copied into the Claude skills directory, or uploaded as an .skill file to Claude.ai. The project is MIT-licensed.
antibody is an open-source coordination layer for fleets of coding agents. It watches tool-call failures, recognizes matching errors across worktrees, lets the first agent claim a diagnosis, and delivers the recorded fix to other sessions using Claude Code, Gemini CLI, and Codex CLI hooks. It does not start agents or manage worktrees. A failed tool call is normalized and converted into a fingerprint so equivalent failures match despite changing ports, paths, timestamps, or identifiers. Claims prevent multiple agents from diagnosing the same new error; once a fix is recorded, it is stored as redacted, human-editable memory and injected into later matching sessions through the host agent's hooks. The project also provides five MCP tools, the `antibody watch` fleet view, statistics, import/export and review commands, and an optional HTTP relay for sharing fixes between clones or machines. Shared memory normally resides in the repository's common Git directory and requires no server or database. The committed Claude Code plugin and the Gemini CLI and Codex CLI setup commands require Node.js 22.13 or newer. The repository is distributed under the MIT license.
big-arrow-on-the-screen is a macOS command-line tool and agent skill for Claude Code and Codex that lets AI agents point at controls without clicking them. Its Swift binary places a click-through overlay above windows and displays a labeled arrow, ring, or box aimed at a target found by coordinates, window or tab, application element, mouse position, or Peekaboo data; the overlay can raise the target application, follow it, speak or provide copyable text, and remove itself after a duration, click, stop command, or agent-process exit. It does not type, click, capture the screen, or contain AI. The project is distributed as a Homebrew install or source-built macOS binary, includes commands for installing its agent skill, and is MIT licensed.
BIGWORDS.PAGE is a static full-screen text display for signage, presentations, kiosks, and other screens. Its editor formats a message to fit the screen, with support for a small Markdown subset, slides, date or duration countdowns, animations, bundled fonts, aspect-ratio framing, images, QR codes, and keeping the screen awake. The viewer stores the display state in the URL fragment after the #, which browsers do not send to the server, so shared displays require no backend, account, or server-side storage; QR codes can open the display on another device. The project can be self-hosted as static files or with its official Caddy-based Docker image. It is written in TypeScript, requires Node.js for development, and is licensed under the MIT License, while bundled fonts and libraries retain their own licenses.
Excalidraw CLI is a command-line tool for managing Excalidraw+ workspaces and rendering local `.excalidraw` files to PNG. Local rendering runs without an account; workspace operations use an Excalidraw+ API key and include listing, creating, renaming, moving, pinning, trashing, reading, replacing and patching scenes, managing collections, and handling workspace members, invites and audit logs. Scene content patches merge elements by ID, use the higher version, and can mark elements as deleted. Commands output JSON by default for use with shell scripts and coding agents, with stdin and stdout support for pipelines. It is distributed as the `@excalidraw/cli` npm package, requires Node.js 22 or newer, uses an installed Chrome, Edge, Chromium or Firefox browser for rendering, and is licensed under MIT.
ff-tracking is an MIT-licensed source project that renders a six-second motion graphic of a virtual camera filming an AI agent's terminal. Its pipeline draws a glyph-positioned terminal in an SVG first pass, exports tracker data, then applies a Skia Metal pass with one SkSL shader for perspective, depth of field, color processing, bloom, an LED grid, scanline tearing, grain, and camera motion. The tracker follows the newest characters, controls the focal plane and overlays, and supplies coordinates to the synthesized soundtrack, which adds hum, keystroke clicks, cut glitches, panned pentatonic lock beeps, and a final chime. The Rust workspace uses fframes, Skia, and FFmpeg; rendering requires macOS with Apple silicon, Rust, FFmpeg, and Python with NumPy and SciPy. An optional Slab-based soundtrack is provided, with Slab required only to re-render that score.
FinderSearch is a native macOS file browser with fast fuzzy filename search, built with SwiftUI and AppKit and powered by the fsearch engine. It searches locally with fuzzy matching, relevance ranking, folder scoping, and filters such as ext:pdf and mtime:<7d, then presents results through icon, list, column, or gallery views with Quick Look previews and tabs. File operations include drag selection, copying and moving, batch renaming, ZIP tools, tags, sharing, Trash, undo, transfer progress, and conflict handling. It requires macOS 15 or later and is distributed as a DMG for Apple Silicon. Full Disk Access is needed to search protected folders; the app is currently not Developer ID-signed or notarized. FinderSearch is early software whose stated unsupported features include content search, AirDrop, Finder extensions, saved smart folders, folder merging, and persistent undo history. It is MIT-licensed, while its vendored fsearch backend is also MIT-licensed.
FSearch is a whole-disk file-search tool for macOS, available as a command-line application with a background daemon and as a Rust crate. It indexes file names and text content, supports fuzzy matching and typo tolerance, and accepts filters for folders, extensions, file types, size, modification time, paths, regular expressions, and symbol names. The index is built by crawling the disk with macOS's getattrlistbulk and kept current through FSEvents; file names are stored in an mmap'd structure, while text search uses a trigram index and reads matching content directly from disk. It exposes JSON Lines over a Unix socket or standard input, and the CLI and Rust API share the same index. The repository reports a 1.3 ms median whole-disk name-search time across 7.7 million files and folders on an M4 Max.
Gentle Dot is a personal AI assistant for macOS that stays available from the desktop, maintains one continuous conversation with long-term Engram memory, reads attached files, and can operate applications through screen-based clicking, typing, shortcuts, and app launches. Each computer-control session requires approval in a native dialog; risky actions require additional confirmation, and a panic shortcut can stop control. The desktop application uses Tauri 2 and a native helper, while a Node daemon runs the Gentle Shell agent engine and exposes its events to the desktop and browser interfaces over WebSockets. It supports voice input and spoken replies, configurable model accounts, MCP-based connectors, and a browser or Docker-hosted server interface. The project is an early preview with no published releases; the code is MIT-licensed and is built by Gentleman Programming on top of Gentle-AI.
good-css is a collection of opinionated modern CSS techniques packaged as an agent skill for coding agents. It guides agents toward CSS declarations that adapt without breakpoint-heavy layouts and toward CSS features instead of JavaScript; the techniques work in plain CSS, Tailwind, StyleX, or other CSS systems. The accompanying PRACTICES.md documents when to use each technique, shows its CSS, explains why it works, and credits its source, while good-css.com provides a live specimen for every entry. The skill can be installed for agents that support the Agent Skills format, including Claude Code, Codex, and Cursor, or added manually; the repository is MIT-licensed.
Habblaud is a local pixel-art dashboard for monitoring Claude Code and Codex sessions. It maps each project to a room and each agent or subagent to a character, displaying live activity, tasks, shell commands, status, session history, usage, and requests for permissions or answers. It reads the session files and transcripts already written on the computer, polling them into a Node server and streaming the office state to a browser over SSE; its canvas renders the office and characters as procedural pixel art. Optional Claude Code plugins and Codex hooks provide live usage data, permission handling, and messages sent back to sessions. The terminal view and actions that affect sessions are restricted to local access, while Docker mounts account data read-only and stores Habblaud data separately. It supports macOS and Linux, can run with Node.js or Docker, and is released under the MIT license.
Helix Foundry is a local-first data workspace built on HelixDB and DuckDB. It connects databases, SaaS tools, REST APIs, object storage, and uploaded files, then proposes an ontology that links matching customers and records across those sources. The workspace stores immutable, versioned Parquet snapshots, profiles schemas and reviews schema drift, and queries the snapshots through isolated DuckDB processes. Its Home and Analyst views provide metrics and answer questions with the SQL executed and citations to the relevant snapshot. AI runs locally by default through Ollama, with Claude and OpenAI also supported; AI-proposed changes require review. The application runs on the user's computer without accounts or hosted infrastructure, and provides an OpenAPI interface, TypeScript SDK, scoped API tokens, and backup and restore scripts. It is distributed under the Apache License 2.0.
A maintained collaborative research repository on conditional integer multiplication bounds, led by Douglas Colkitt and based on the OpenAI #109 framework. It records community constructions, proof arguments, exact certificates, contribution history, and reproducible validation checks. The current construction combines a shared-edge complex supplier with a physical bit-word and searched modules, coordinates operation frames, completes shared entrance banks, and supplies a conflict-free bank schedule with finite prime exclusions. Its asymptotic bound and all-size interfaces remain conditional on analytic, uniform-recursion, and fixed-tape hypotheses; the repository states that finite replay and written review do not formally verify the complete multiplication theorem or demonstrate a practical runtime speedup. The project is licensed under Apache-2.0, with separately licensed bundled sources and preserved attribution notices.
jactionlint is a static checker for GitHub Actions workflow files, an actively maintained fork of rhysd/actionlint. It checks workflow syntax, types in `${{ }}` expressions, action inputs and outputs, reusable workflow interfaces, composite actions, and dependabot.yml files. It integrates ShellCheck and Pyflakes for scripts run by workflows and includes security and policy checks for script injection, unpinned actions, excessive permissions, dangerous triggers, cache poisoning, hard-coded credentials, and missing timeouts. The tool provides correctness, default, and pedantic profiles, automatic fixes with `--fix` or `--diff`, baselines for adopting checks incrementally, durable ignores, and SARIF, GCC, and GitHub output formats. Optional online checks cover impostor commits, vulnerable actions, archived repositories, and stale references. It runs locally or in CI, and is distributed as a binary, a Go package, a Docker image, and through package managers including mise and Homebrew. jactionlint is distributed under the MIT license.
Liquid Glass is an open-source library of glass surfaces and controls for React, the web, and Electron, developed by Glass-HQ and built with vgpu. Its Clear and Regular materials support light and dark appearances and refract live content; the package provides components such as GlassScene, GlassContent, and GlassButton for wrapping interfaces in a glass scene. It is distributed through npm as @glass-sdk/liquid-glass under the MIT license. Rendering requires HTTPS or localhost and an available WebGPU adapter, and the project notes known rendering issues in Safari.
Lithos Metal is an open-source LLM inference engine for Apple silicon that generates specialized Metal megakernels and serves models locally for coding agents and other applications. Its compiler decomposes operators into matrix tiles, attention partitions, and recurrent-state slices, records their dependencies, and generates layer-wise fused GPU tasks. The runtime combines mixer megakernels, GPU-resident generation control, pre-encoded Metal indirect command buffers, and DSpark speculative decoding, in which a target-conditioned draft proposes token blocks and the target model verifies and commits the accepted prefix. The server provides OpenAI-compatible Chat Completions and Responses adapters plus an Anthropic Messages adapter, with text generation, tool calls, and SSE streaming; launchers configure clients including OpenCode, Claude Code, Codex, and Hermes. It runs on Apple silicon with macOS 26 or later, is distributed through a Lithos Homebrew tap or source installation, and is licensed under Apache 2.0.
omarchy-apple-dev is an open-source toolchain for building and deploying iOS apps from Omarchy Linux on x86_64 or Apple Silicon PCs, without macOS or an installed copy of Xcode. It supports Swift, SwiftUI, and Flutter projects, using Swift, xtool, pymobiledevice3, Linux replacements for selected Xcode tools, and an iOS SDK extracted from Apple's Xcode distribution. Apps can be built and installed on a physical iPhone over USB; free Apple IDs are sufficient for device installs, while TestFlight and App Store distribution require a paid Apple Developer account and App Store Connect credentials. The repository includes scripts for installing the toolchain, pairing and launching apps with LLDB, adapting Xcode projects for xtool, signing and packaging applications as IPAs, validating bundles offline, and uploading builds to App Store Connect. Flutter support is limited to release builds, with no debug, hot reload, or simulator support. It is licensed under the MIT license.
omdrop-owl is a Linux radio backend that brings AirDrop communication to non-Apple devices without requiring Apple's Broadcom Wi-Fi hardware. It uses OWL, a user-space implementation of Apple's AWDL protocol, on a monitor interface while the Wi-Fi station remains connected; the omdrop sender and receiver communicate through the resulting awdl0 interface. The backend is designed first for the MediaTek MT7925 and advertises AirDrop services through the omdrop tools. It shares the access point's channel, which must be 6, 44, or 149, and plugs into omdrop through helpers installed under /usr/lib/omdrop. The repository includes build and packaging support for Arch, Debian, and Ubuntu, along with hardware profiles and a card-testing script. The MT7925 profile is tested for sending and receiving while Wi-Fi remains connected; Intel AX200 and AX211 support is documented as nonfunctional.
once is a command-line tool that runs a command once and reuses its standard output for a specified period. It keeps successful results in memory in a per-user background daemon, keyed by the tenant, working directory, command, and arguments; repeated calls return the cached output until the TTL or an optional absolute expiry is reached. The daemon stops when its last entry expires. The command runs in the current directory with the caller's environment. Standard input and standard error pass through, non-zero results are not cached, and --no-dir can omit the working directory from the cache key. Tenant keys separate cache contexts, while status and clear commands inspect or remove cached entries. A documented use case is caching a 1Password CLI secret after the first fingerprint approval. The project is distributed as a Go program with no dependencies beyond the standard library and supports macOS and Linux. Cached values remain unencrypted in daemon memory; access is restricted to processes of the same user. Outputs larger than 64 MiB are passed through without caching. It is licensed under the MIT License.
Open Intelligent UI is an open-source implementation of ChatGPT's interactive-answer experience, built on OpenUI. It renders streamed answers as custom travel interfaces rather than plain text, including photographs, an editable itinerary, and a MapLibre street map with pins, filters, route framing, and a route line. The model streams OpenUI Lang through OpenUI Gateway, which validates and corrects the output against the repository's component library. Travel components include headings, prose, image galleries, stops, itineraries, maps, and route suggestions; pins appear during streaming, and adding or removing stops updates OpenUI response state so the changes are sent to the model on the next turn. Gateway's image-search tool supplies photos, with Wikipedia fallbacks, while server routes cache Wikipedia lookups and OSRM driving-route geometry. The repository contains a runnable Node application requiring Node 24. It can use OpenUI Gateway with a THESYS_API_KEY or run a local model through Ollama; local mode has no Gateway image search or output correction and uses Wikipedia photo fallbacks. The project is independent of and not affiliated with OpenAI.
OSS Scanner is an Anthropic service that scans enrolled critical open-source repositories for security vulnerabilities. Maintainers enroll a project by opening a pull request that adds a project directory with a configuration file and Dockerfile; the Dockerfile sets up dependencies and builds the project, initially with network access enabled. After the build, the project is moved into an isolated virtual machine with Internet access disabled for the security analysis. Findings are emailed to the configured contacts with reproduction steps and, where available, a proposed patch. Reports are model-generated and are not reviewed by a human. The repository includes validation and local-checking tools for testing the build and no-network audit environment before enrollment.
A GitHub repository containing a device-specific guide and scripts for taking control of locked-down Android photo frames, including the Skylight Calendar 15-inch model and the BIUFRAME 10.1-inch frame. It documents a nine-phase process covering hardware identification, ADB access, loader mode, verified firmware backups, rooting, custom launchers and kiosk applications, networking, vendor updates, and security hand-off. The repository also includes a Vite family-dashboard web app deployed to Cloudflare Workers and a small full-screen Android WebView kiosk app, with per-device notes, scripts, research documents, and explicit rules for approving and verifying flash writes. It warns that frames sold under the same name can use different boards and firmware, so its procedures are not universal flashing guides.
PiDurableKit is a Swift package for running durable AI agents on iPhone, iPad, Mac, and Vision Pro. It embeds the pi-durable, pi-ai, and chord JavaScript libraries in JavaScriptCore and exposes them through a native Swift concurrency API. Conversations, model turns, tool calls, tasks, and application state are committed to persistent storage before being exposed to the app, allowing unfinished work to resume after a crash or an iOS process termination. The package supports SQLite, JSONL, or in-memory storage; Swift and JavaScript extensions; tools, hooks, prompt sections, subagents, durable checkpointed tasks, transactions, typed documents and entries, streaming conversation views, and coding tools in a sandboxed directory. It supports pi-ai providers including Anthropic, OpenAI, Google, OpenRouter, Groq, xAI, and Mistral, as well as compatible servers and OAuth sign-in with credentials stored in the Keychain. It requires iOS 17, macOS 14, tvOS 17, or visionOS 1. PiDurableKit's own code is dedicated to the public domain under CC0 1.0; its bundled dependencies use the licenses listed by the project and require their notices to be shipped with applications.
REA is a local reverse-engineering tool for AI agents, distributed as a CLI and MCP server. It connects agents to analysis providers such as Hopper and Ghidra, and can inspect native binaries, macOS applications, JavaScript and Electron applications, .NET assemblies, websites, packages, and controlled runtime behavior. Its investigation workflow decompiles or maps a target, follows strings, symbols, calls, references, and control flow, records evidence and limitations, and uses the findings to help reconstruct features in another project without claiming to recover original source code or automatically clone an application. Analysis runs on the local host; REA configures agent integrations and can optionally install Hopper, while Ghidra and Java must be installed separately. The project is distributed under the MIT license.
ref2game is an MIT-licensed skill from studioigor that turns a reference image or concept into a playable 2D browser-game prototype for Claude Code or Codex. The agent studies the image, asks setup questions, creates a project skeleton and style/layout plan, generates art piece by piece, and develops a static playable build while comparing frames with the reference. Its runtime uses a small WebGL2 engine in which code controls rigs, two-bone character animation, walk cycles, wind, lighting, particles, camera behavior, HUD elements and other effects rather than baking them into the generated images. Background critics and automatic checks evaluate assets, layout, fidelity, motion, foot sliding, joint behavior, collisions and other reported defects; feedback is converted into code changes or regenerated assets and a new check. The final output includes a playable build, textures, manifest and source files. The skill has separate editions for Claude Code and Codex, supports image generation through Codex, image APIs, local tools or manual assets, and requires Claude Code or Codex, Python 3.9 or later with Pillow, NumPy and SciPy, and Node.js 18 or later. It supports macOS and Linux; Playwright with Chromium is installed on first run, while ffmpeg is needed for video references and renders.
replica-skill is a set of eleven MIT-licensed Claude skills for researching and rebuilding an existing app. The skills reverse-engineer public pages, screenshots, store listings and a user's own account into screens, flows, components and an inferred data model; plan the stack, database schema and API; reconstruct the design system and application; add authentication, payments and integrations; test flows with Playwright; compare the result with the original; analyze linked user reviews for complaints; rebrand the clone; create launch materials; and deploy it. The skills pass work between stages through files in a project's replica/ directory and can be installed in Claude Code as a plugin, copied into Claude's skills directory, or run by pasting a SKILL.md into a chat. Included Python utilities use the standard library and cover screenshot diffing, feature-parity scoring, review analysis, contrast checking, codebase-name cleanup and store-listing checks. The project states that it rebuilds functionality and user flows rather than source code, proprietary assets, branding, copy, content or private APIs, and that it uses public sources and the user's own account without bypassing logins, paywalls or site terms. It requires Python 3.8 or newer, has no package-install requirement, and is distributed under the MIT license.
SeeCode is an open-source agent skill that turns short JSON specifications written by Claude and other coding agents into animated, explorable diagrams. Its renderers handle layout, edge routing, labels, typography, accessibility and motion rather than requiring the model to write SVG; diagrams can be explored as self-contained HTML files and exported as SVG, PNG, JPEG, GIF or MP4. It supports architecture, flow, sequence, state, data, isometric and chart diagrams, with data loaded from CSV or JSON for charts. SeeCode can import Mermaid, Graphviz, PlantUML, D2, draw.io, Excalidraw, Structurizr, BPMN, SQL, Prisma, DBML, OpenAPI and common data files, and can scan a repository to link diagram nodes to source files. It works with Claude Code, Cursor, Codex, OpenCode, GitHub Copilot and other Agent Skills hosts, requires Node 20 or later for its command-line tooling, and is licensed under the MIT License. The project states that it has no server, accounts or telemetry.
Speakrail is a fully local, full-duplex voice assistant that runs on a single 24 GB NVIDIA GPU such as an RTX 4090. It combines Voxtral Mini 4B Realtime for speech recognition with a turn-taking head that evaluates the conversation every 80 ms, and uses Gemma 4 12B with a Speakrail LoRA to decide whether to speak, listen, interrupt, yield, or interject. Breeze TTS 2 produces the assistant's speech; peek decoding can begin preparing a response before the turn is fully confirmed, while listening notes preserve longer or corrected requests. The local web application provides tools for weather, timers, notes, lists, calculations, unit conversion, and optional web search or external model access. It is distributed as Docker Compose services and requires Linux, Docker with NVIDIA Container Toolkit, roughly 75 GB of disk space, and a suitable NVIDIA GPU. Speakrail's code is licensed under Apache 2.0, while the default Breeze TTS 2 model and generated audio are restricted to research and non-commercial use.
Strands Box is an open-source sandbox engine for running AI agents. Written in Rust, it combines operating-system isolation with Dogwood semantic policies: direct file, program, and network access is configured in box.toml, while operations handled through Strands Shell, Monty for Python, the egress gateway, and the MCP broker are evaluated by an embedded Dogwood Local Engine. Operations are denied by default unless a permit rule allows them, and forbid rules override permits; rules can inspect arguments, earlier actions, and elapsed time, allowing activity in one interpreter or tool path to affect later requests. Box also supports credential injection for permitted requests, keeping configured API credentials or AWS SigV4 signing secrets outside the agent, and records policy decisions as OTLP JSON. The repository describes the current preview as local execution on macOS with Apple silicon, with Linux support planned, and distributes the project under the Apache License 2.0.
TraceMaker is an open-source autorouter and component placer for KiCad circuit boards. It reads a .kicad_pcb file, routes unrouted connections by drawing tracks and vias, and writes the result back to a KiCad file; its output is checked with KiCad's own design-rule checker. A live browser viewer can display routing progress, and the separate placer can improve an existing placement, steer placement for routability, nudge selected parts, or place a board from scratch while respecting locked parts, edge connectors, and keep-outs. The project provides a C++20 command-line application, a KiCad 10 action plugin, a design-rule checker, inspection and package-escape commands, and CPU, CUDA, and Apple Metal execution paths. It runs without a GPU; CUDA 12.x or Metal can accelerate parts of the computation. The repository reports support for KiCad 9 and 10 boards and includes benchmark and test tooling. It is released under GPL-3.0-or-later.
TradingChart is a native SwiftUI trading-chart package for iOS, built on Swift Charts. It renders candlestick, line, and area series and supports live updates, scrolling and pinch zoom, crosshair tooltips, paged history, buy and sell markers, theming, localization, and horizontal, trend-line, and ray drawing tools. Its indicator modules provide SMA, EMA, WMA, Bollinger Bands, Volume, RSI, and MACD overlays or panes, and a ChartIndicator protocol allows custom indicators. The package separates the data model and math into TradingChartCore, indicator calculations into TradingChartIndicators, and the SwiftUI chart view into TradingChart; the main product re-exports the two lower-level modules. ChartPersistence can save styles, indicators, and drawings, while ChartSeries validates and organizes price points or candles by interval. The chart UI requires iOS 17 or later, while TradingChartCore and TradingChartIndicators support iOS 16. It is distributed as a Swift Package, is licensed under the MIT license, and includes a demo app and unit-tested core modules.
VB6 Studio Web is a browser-native classic Visual Basic development environment, form designer, source runtime, and controls library built with HTML, CSS, and JavaScript. It lets users edit VB-style source, design forms, run projects, step through them in a debugger, open Visual Basic 6 .vbp projects and groups, and export projects as independent single-file HTML applications. Its compiler and cooperative virtual machine execute VB-style source through their own instruction model rather than JavaScript eval. The IDE includes project and source editing tools, classic controls, events, data-source adapters, MDI support, conditional breakpoints, statement stepping, watches, error trapping, and debugger inspection. Rendering uses WebGPU with a Canvas2D fallback, and the project includes browser-compatible subsets of Win32 and GDI functionality. The project requires Node.js 22 or later, has no npm package dependencies for its main build, and can run from a local server or static host without a service, database, or account. It is MIT-licensed. The repository states that it is not a complete, native, or pixel-identical Microsoft VB6 replacement; native COM/OCX execution, full VB6 and Win32 compatibility, and some designers and APIs remain limited or unfinished.
Searchable transcript of GitHub Trending Today #53: omarchy-apple-dev, replica-skill, big-arrow-on-the-screen, fsearch, once — Github Awesome (14:56). Search for a phrase, then click its timestamp to jump straight to that moment in the video.
Captions sourced from the original video on YouTube, published by Github Awesome. The video, its captions and all related intellectual property remain the property of their respective owners; AINotes claims no ownership. Provided for research, accessibility and search — see the Transcript Notice and Copyright Policy.
00:00 Welcome back to GitHub awesome. This is GitHub trending today number 53. 35 trending open- source projects on GitHub right now. Let's get into it. Omari Apple-Dev lets you build Swift UI and Flutter iPhone apps from Omar Linux, then install them on a real iPhone over USB. There's no Mac OS or Xcode installation in the workflow, though the full setup still needs Apple's Xcode download for its iOS SDK.
00:26 It also packages and validates apps for test flight or the app store with a paid Apple developer account required for those releases. Replica skill is 11 clawed skills for building your own take on an existing app. It maps the screens and user flows, plans the back end, builds and tests the result, then compares what's missing. One clever step reads linked user reviews to find complaints your version could address.
00:54 The workflow uses public information or your own account, writes fresh code, and requires new branding before launch. REIA gives coding agents a way to investigate software when the source code isn't available. It connects through MCP to inspect native binaries, Electron apps, Android packages, websites, and more. Then returns evidence your agent can follow up on.
01:19 Ask how a feature works, and the agent can trace the relevant pieces before building its own version. The analysis runs locally with GIDRA, Hopper or IDA needed for deeper native work. Big arrow on the screen solves a tiny but familiar AI agent problem. It needs you to click something and its message is buried in a terminal. On a Mac, the agent can draw a labeled arrow, ring, or box over the exact control you need.
01:45 Your clicks still reach the app. Keyboard focus stays put and the pointer disappears on its own. It's a simple visual handoff for permissions, setup, and step-by-step help. FS search is a Mac OS command line search tool for that file you know exists but can't quite name. It indexes the disk, forgives typos, and searches inside text files with filters for folder, type, size, and date.
02:07 The author reports a median 1.3 milliseconds for name searches across 7.7 million files on an M4 Max. A small Damon keeps the index current as files change. Omdrop Owl brings AirDrop to Linux machines that don't have Apple's Wi-Fi chip. It's the radio half of Omdrop built on Owl, a user space reimplementation of Apple's ADDL protocol. The trick is keeping your Wi-Fi connected while AirDrop runs on the same channel by injecting frames from the station's MAC and letting it act the phone's traffic so you can send and
02:44 receive from an iPhone without dropping your network. West Boss took two locked down Android picture frames, a skylight calendar and a BYU frame and turned them into proper dashboards. The repo logs how he got ADB, backed up the firmware, rooted them, pulled out the vendor app, and ran his own launcher and kiosk on the Skylight. Root came from adding one file to the boot RAM disc.
03:07 The guide is written partly for AI agents as a nine-phase map an agent can follow over USB. AI homepage replaces Chrome's new tab with a homepage an AI agent writes for you from your own browsing. It ranks the sites you visit most, opens up to eight of them in a background window signed in as you, pulls out the headlines and links, then writes a single React component that becomes your new tab.
03:35 Tell it what you want in plain English, a morning newspaper or a green on black terminal, and it rebuilds on a timer. Helix Foundry is a local data workspace that connects your company's scattered tools, Postgress, Stripe, Postthog, S3, and more. Then proposes one connected ontology across all of them. It snapshots your data as version parquet, queries it with duct DB, and stores the relationships in Helix DB.
03:59 Ask the analyst a question, and it answers with the SQL it actually ran, citing the snapshot. OSS Scanner is Anthropic's free service that hunts for security bugs in open-source projects and maintainers sign up with a single pull request. You add a folder with a config file and a Docker file, plus an optional threat model describing what you actually care about.
04:24 The scanner builds your project in an isolated VM, then cuts its internet access before it starts looking. CCode lets Claude and other coding agents draw proper diagrams without ever writing SVG themselves. The agent writes a short JSON spec of what connects to what and CCO handles the layout, edge routing, labels, and animation, then outputs one self-contained HTML file.
04:49 There are 42 diagram types from architecture and sequence to seni and wardly maps. Speak rail is a fully local voice assistant that listens while it talks running on one RTX 4090. You can cut it off mid-sentence, say m without it stopping, or pause to think without it jumping in. Every 80 milliseconds, a small turnt head decides whether you're done, pausing, or just back channeling.
05:15 And Gemma picks speak, listen, or yield in about one token. Finder search is what happens when a lightning fast file name search engine gets a familiar Mac interface. Start typing even with a typo and it finds matching files across your disc. From there you can browse in columns, icons, lists or a gallery. Preview with quick look and handle files without opening another app.
05:40 It uses F search under the hood. Keep searches on your Mac. FF tracking is the source for a six-second motion graphic that makes an AI agent's terminal look like it's being filmed by a restless camera. A tracker follows the letters as they appear. The focus shifts with it, and each lock gets a synthesized beep. There's no recorded footage or stock audio.
06:02 The screen, camera effects, and sounds all come from code. Change the text or camera moves, then render your own version. Jackson Lint checks GitHub actions workflows before a broken or risky one reaches CI. It catches bad YAML keys, mistyped expressions, wrong action inputs, and scripts that drop untrusted text straight into a shell command. It also flags issues like overly broad permissions, and unpinned actions.
06:29 Run it across a repo, use its suggested fixes, or set a baseline so an existing project only fails on new problems. Strandsbox is a sandbox for running AI agents with two layers of control. The OS walls off files and programs and a policy language called Dogwood decides every shell command, Python call, network request, and MCP tool call with everything denied unless a rule permits it.
06:56 Because the gateways share a history, a policy can say once the agent has read a secrets file, block its next outbound request. Good CSS is a set of modern CSS habits packaged for coding agents with a live example for each one. It nudges an agent toward fluid sizing with clamp color tokens that handle light and dark modes and layouts that adapt without piles of breakpoints or JavaScript.
07:20 Each practice says when to use it and shows the actual CSS so the advice can fit a plain stylesheet, tailwind or another setup. AI data extractor pulls your chat history out of 10 AI coding assistants, claude code, codecs, cursor, and more into one normalized JSONL format. It finds where each tool stores its chats on Mac, Linux, or Windows, reads SQLite databases readon, so a running editor isn't blocked, and keeps the code context, diffs, and tool calls alongside each message.
07:54 Absent author gives researchers and reviewers a way to check whether anyone really stood behind a paper. Its two agent skills help authors map claims to evidence and help reviewers find concrete problems like references that don't exist or numbers that change between tables. A separate llinter flags writing patterns for human inspection. The important restraint polished pros or an AI looking phrase is never treated as proof of research.
08:22 Open intelligent UI is an open recreation of chat GPT's interactive answers. Ask for a one-day San Francisco plan and you don't get a wall of text. You get a map, photos, and an itinerary you can edit. Pins drop onto the map while the answer streams. Then the route draws in once it finishes. Remove a stop or add a suggestion and your edits go back to the model on the next turn.
08:44 Once is for commands whose output you need again, but don't want to fetch again. wrap a command, set an expiry, and later calls reuse its output from a small background process. The handy example is reading a one password token. Approve it with your fingerprint once, then keep using it for the next few hours. Cached values stay in memory, and the process exits when they expire.
09:11 Purable kit lets an iPhone or Mac app run AI agents that survive being killed. It wraps the PI durable agent library and runs it in JavaScript core behind a native Swift API, writing every conversation, model turn, and tool call to storage before the UI shows it. So when iOS kills your app mid turn, you reopen it and the agent carries on from where it stopped.
09:35 Isoigure is a clawed skill that draws a working isometric object in a single HTML file. Ask for a kitchen timer or a light switch and you get a fine line technical drawing on a numbered plate where the buttons actually do something when you press them. It's plain SVG with no libraries. About 15 lines of projection code put every face on its isometric plane.
09:59 Antibbody tackles a familiar problem when you run several coding agents. They all hit the same error and each spends time solving it. It recognizes matching failures across work trees, lets one agent claim the diagnosis, then passes its recorded fix to the others. The shared memory lives with the repo and it works with claude code, Gemini CLI, and Codeex CLI.
10:19 One fix, fewer agents stuck on the same wall. VB6 Studio Web brings the classic Visual Basic workflow into your browser. You can lay out forms, edit VB style code, run it, step through it in a debugger, and export a project as a standalone HTML app. It can also open original VB6 project files. This is a browser implementation with compatibility limits, but it's a surprisingly hands-on way to explore and adapt old Visual Basic projects.
10:51 Liquid Glass brings that glassy refracting look to React and the web with real live refraction rendered on the GPU through web GPU. You get two materials, clear, which is fully transparent, and regular, which is frosted. Both come in light and dark. Wrap your surfaces, buttons, toolbars, tabs, and menus in a glass scene, and they bend whatever sits behind them.
11:11 Ref 2 game is a skill for turning a reference image into a playable 2D game prototype. Give Claude code or codeex a screenshot and it guides the agent through matching the art, building a browser game, and adding movement, lighting, and effects in code. You can play the work in progress, point out what feels wrong, and get a static build with the source and assets.
11:38 Bigwords.page puts your message full screen in giant text that fits any display with nothing else on the page. Type it in the editor and the whole message lives in the link itself. after the hash, which browsers never send to the server. So there's no backend, no account, and nothing stored. Separate slides with two pipes add a countdown to a date or drop in a QR code for your Wi-Fi.
12:03 Tracemaker takes a Kyad circuit board with unfinished connections and draws the tracks and VAS for you. It can also move components to make a board easier to route. And there's a live viewer so you can watch the paths appear. The useful check is Kyad's own design rules, not just the router saying it succeeded. It runs on the CPU with optional GPU acceleration.
12:27 Lithos Metal is an open-source engine for running large language models locally on Apple Silicon. It builds specialized Metal kernels to cut GPU overhead, then serves the model through APIs that coding agents can use. There are launch commands for codeex, claude code, and open code. The current work centers on specific Quen models with performance measured on selected Macs, so check its hardware list before trying it.
12:53 Excaladraw CLI puts your drawings in the terminal. You can render a local.exal file to a PNG without an account, which makes diagrams easy to generate in a script or attach to a build. With an Excaladraw plus API key, it can also list, create, and edit scenes and collections in your workspace. Commands return JSON by default, so coding agents can work with them, too.
13:19 Integer malt bounds asks a surprisingly deep question. How much faster can we multiply enormous integers? Contributors are sharpening a mathematical bound built on an earlier open AAI research framework with proof arguments, exact certificates, and checks others can rerun. The latest reviewed result improves the claimed bound, but it still depends on assumptions in that framework.
13:43 Habla turns your clawed code and codec sessions into a little pixel art office. Each project gets a room and each agent becomes a character you can watch working, waiting on a command, or raising a hand when it needs you. Click one to see its activity and session details. It runs locally, making a crowded set of agent tasks much easier to understand at a glance.
14:07 Trading Chart is a native Swift UI library for iOS trading charts built on Apple's Swift Charts. You get candlesticks, line, and area styles, plus SMA, EMA, Ballinger bands, RSI, and MACD with a protocol for writing your own. Swift Charts draws the bars, while lightweight canvas layers handle the axes, crosshair, and drawings. So, the chart isn't rebuilt on every scroll frame.
14:29 Gentle Dot is a personal AI assistant that lives on your Mac desktop, one key away. It keeps one continuous chat with long-term memory, reads your files, and can drive your apps, seeing the screen, clicking, typing, and opening things. Each control session needs your approval. Risky actions ask first. Password managers and system settings are off limits.