theme: channing-cyan highlight: atom-one-dark

Open Source Growth Pulse Today

2026-10-29 · 20 projects · 24h Star increment tracking

This issue selects 20 open source projects worth attention based on their 24-hour star increment. Refer to each repository for exact capabilities.

Today in 30 Seconds

Rank	Project	24h Increment	Language
🥇 1	morluto/rea	+5,426	TypeScript
🥈 2	Ebony-Vinyl/dsh-our-free-model	+2,203	JavaScript
🥉 3	Gathin23/cleanyourtv	+2,187	Python
4	pbakaus/impeccable	+2,049	Rust
5	HKUDS/DeepTutor	+2,046	Python
6	Nutlope/hallmark	+1,947	CSS
7	interplains/Zero-Breach-Toolkit	+1,831	Python
8	debpalash/VoiceStudio	+1,738	Python
9	hmjz100/LinkSwift	+1,659	JavaScript
10	elayadesign/ai-design-skills	+1,652	Unknown
11	cloudflare/security-audit-skill	+1,594	JavaScript
12	google/elephant-goldfish-model	+1,582	Python
13	unclebob/ub-mode	+1,519	Unknown
14	KnifeLemon/Filee	+1,498	C#
15	togg53192-cmd/jailbreaks	+1,466	Unknown
16	AienTech/dossier	+1,372	TypeScript
17	seedds/nxnative	+1,299	Objective-C
18	ifixai-ai/iFixAi	+1,285	Python
19	anhao/ai-pose-gallery-zh	+1,179	HTML
20	Zagitta/gpu-unlock	+1,136	C

Today's largest increment: +5,426 (morluto/rea) · Top three combined: +9,816

Project Deep Dives

🥇 01｜rea ⭐ 402,326 (24h +5,426)

REA is a reverse engineering tool based on TypeScript that provides a unified MCP (multi-function plugin) framework for analyzing binary files, applications, and runtime behavior. Through pluggable agent modules, it helps users quickly locate feature implementations, understand internal mechanisms, and can be used directly across multiple platforms (such as Hopper). It suits security researchers, software reverse engineers, and developers, addressing the challenge of moving from high-level functionality down to low-level binaries.

Project page: github.com/morluto/rea

🥈 02｜dsh-our-free-model ⭐ 48,844 (24h +2,203)

dsh-our-free-model is a plugin for the dsh platform that provides free, unlimited access to cutting-edge models such as DeepSeek V4.1 Flash and Kimi K3, without login, registration, or an API key. It mounts directly with zero dependencies and no build steps, and suits users who need to run high-performance LLMs quickly in local or private environments.

Project page: github.com/Ebony-Vinyl/dsh-our-free-model

🥉 03｜cleanyourtv ⭐ 43,788 (24h +2,187)

tv-vibehack is an ADB-based tool that helps users remove ads and preinstalled junk from Android TV/Google TV without root or physical connection, and install a clean launcher that looks like Apple TV. It suits ordinary home users and tech hobbyists.

Project page: github.com/Gathin23/cleanyourtv

04｜impeccable ⭐ 92,544 (24h +2,049)

Impeccable is a frontend design guidance system for AI programming assistants. With 23 commands, 59 deterministic detection rules, and real-time browser iteration, it helps AI-generated frontends break away from homogeneous design templates (such as Inter font, purple-blue gradients, nested cards). It provides a structured design language and quality review capability for developers and AI coding tools (Cursor, Claude Code, GitHub Copilot, etc.).

Project page: github.com/pbakaus/impeccable

05｜DeepTutor ⭐ 47,906 (24h +2,046)

DeepTutor is a lifelong personalized teaching platform based on Python and Next.js. By integrating multiple large language models, knowledge-base retrieval (RAG), and intelligent teaching agents, it offers personalized tutoring, knowledge management, guided learning paths, and multimodal interaction, aimed at educators and self-learners who need adaptive learning support.

Project page: github.com/HKUDS/DeepTutor

06｜hallmark ⭐ 40,653 (24h +1,947)

Hallmark is a design skill package for AI code editors (Claude Code, Cursor, Codex). It generates AI-trace-free UI code through 57 anti-pattern checks and 21 theme systems. It automatically picks macro structure, color scheme, and typography per request, ensuring each generated page has a unique design fingerprint rather than templated output.

Project page: github.com/Nutlope/hallmark

07｜Zero-Breach-Toolkit ⭐ 2,717 (24h +1,831)

Zero Breach Toolkit is a free open-source offline tool that uses Python to detect traces of known commercial spyware (such as Pegasus, Predator) in backups from Windows, Mac, Android, and iPhone. It reads only and never modifies the system, reports results and evidence strength. It suits journalists, activists, and human rights organizations that need an initial self-check of security.

Project page: github.com/interplains/Zero-Breach-Toolkit

08｜VoiceStudio ⭐ 92,934 (24h +1,738)

VoiceStudio is an open-source, fully local voice cloning and design tool. It supports voice cloning in 646 languages, voice design, video dubbing, dictation, transcription, and audiobook production. It includes a desktop Electron app and local API, aimed at creators, developers, and enterprises.

Project page: github.com/debpalash/VoiceStudio

09｜LinkSwift ⭐ 34,477 (24h +1,659)

LinkSwift is a JavaScript tool for getting direct download links from cloud drives, adapted from "Cloud Drive Direct Link Download Helper". It supports eight major Chinese cloud drives: Baidu, Aliyun, China Mobile, Tianyi, Xunlei, Quark, UC, and 123. It helps individual users quickly obtain direct file links for batch or manual downloads.

Project page: github.com/hmjz100/LinkSwift

10｜ai-design-skills ⭐ 11,064 (24h +1,652)

AI Design Skills is a library of design skills usable by AI coding tools (such as Claude Code, Cursor, Codex, Windsurf). Each skill is stored in its own folder; users copy the corresponding SKILL.md rules file as needed, letting AI complete design tasks such as building a landing page from scratch. It suits developers and designers who need to quickly generate or upgrade web designs.

Project page: github.com/elayadesign/ai-design-skills

11｜security-audit-skill ⭐ 51,616 (24h +1,594)

security-audit is a code agent skill for internal use at Cloudflare that provides a multi-stage, machine-readable security audit process. It runs six independent phases (recon, coverage-driven hunting, candidate validation, structured output, independent record validation, and target-neutral reporting) to generate verifiable finding files. It suits development and security teams that need systematic and traceable vulnerability discovery.

Project page: github.com/cloudflare/security-audit-skill

12｜elephant-goldfish-model ⭐ 4,687 (24h +1,582)

Elephant‑Goldfish Model (EGM) is an AI plugin open-sourced by Google. It helps developers, before coding, collect project context and draft a design specification using the "Elephant" role, then the "Goldfish" role reviews the draft with zero context, revealing underlying assumptions and blind spots. It suits feature development that needs design decisions made early.

Project page: github.com/google/elephant-goldfish-model

13｜ub-mode ⭐ 3,712 (24h +1,519)

ub-mode is a set of agent skills for writing and modifying code that follow strict software engineering principles. Through a three-phase process of "make it work, make it right, keep it working," combined with deterministic quality gates like CRAP, duplication, and mutation testing, it helps developers improve code quality and maintainability. It suits individuals or teams needing rigorous code quality control.

Project page: github.com/unclebob/ub-mode

14｜Filee ⭐ 16,277 (24h +1,498)

Filee is a free open-source offline file converter supporting more than 180 formats, including images, PDF, Office documents, ebooks, audio, and video. Users drag a file onto the cursor and pick the target format from a ring that pops up; all conversion happens locally with no extra software or network upload.

Project page: github.com/KnifeLemon/Filee

15｜jailbreaks ⭐ 7,036 (24h +1,466)

This project collects system prompts, role definitions, workspace policies, and agent operation specifications for 31 jailbreaks targeting 25+ models, covering free invisible models, open-weight models, and closed frontier models. It helps researchers and developers explore how prompt levels, role frameworks, and policy overrides affect model behavior, and strengthen AI system resistance to prompt injection and instruction manipulation.

Project page: github.com/togg53192-cmd/jailbreaks

16｜dossier ⭐ 2,740 (24h +1,372)

Dossier is a self-hosted job application management app. It generates tailored resumes and cover letters for each position from a single profile, and tracks the entire application flow from draft to hire. It supports local or cloud AI models and stores data in your own PostgreSQL database, emphasizing privacy and control. It suits individuals or small teams who want to manage career information on a private server and automate job application documents.

Project page: github.com/AienTech/dossier

17｜nxnative ⭐ 8,985 (24h +1,299)

nxnative is a tool that converts Nintendo Switch game NSP packages into unsigned IPAs so they can run natively on iPhone, iPad, and Mac. It uses NSO rewriting, a Horizon HLE host, and an NVN-to-Metal graphics layer to execute game code directly on Apple silicon, solving the need to run Switch games on iOS/macOS. It targets players who possess legitimate game files and prod.keys.

Project page: github.com/seedds/nxnative

18｜iFixAi ⭐ 37,047 (24h +1,285)

iFixAi is an independent audit tool for AI Agents. It runs 49 checks in 120 seconds to verify that an Agent executes tasks correctly according to business KPIs and org structure. It offers three run modes: CLI wizard, explicit commands, and plugins. It supports cross-vendor evaluation to keep audit results citable, and suits developers and enterprises that need to quickly validate Agent behavior compliance.

Project page: github.com/ifixai-ai/iFixAi

19｜ai-pose-gallery-zh ⭐ 1,846 (24h +1,179)

AI Pose Gallery Chinese version is a zero-dependency responsive static webpage. It offers 1,439 entries and 288 sets of AI human body pose reference images (5 viewpoints each), all localized in Chinese and directly downloadable. It serves as creative reference for AI painting, 3D modeling, photography, and more.

Project page: github.com/anhao/ai-pose-gallery-zh

20｜gpu-unlock ⭐ 12,339 (24h +1,136)

GPU unlock is a Linux kernel module targeting Nvidia RTX 3000 series cards. It unlocks the instruction issue rate for FP/BF16 instructions before driver binding, boosting performance to pro-card levels. It targets researchers, deep learning, and high-performance computing users who need high compute throughput.

Project page: github.com/Zagitta/gpu-unlock

---

Data source: YourTools GitHub Trending, updated daily.

Original URL: https://tools.wflynn.cn/aitools/github-trending

Follow the official account "YourTools" for daily GitHub trends and open-source tool alerts: