← Overview
GitHub Trending · Wednesday

18 November 2026

15 repos rising that day. Updated

Claude Codeagentskill bundle15 commandsfindingstriagepatterns681 reports
+1K that day★ 9.4KPython

Claude Code skill bundle for bug hunting and red-team work

A skill bundle for Claude Code with 15 slash commands and 681 disclosed-report patterns. It targets bug hunting and external red-team work.

  • Patterns are curated from disclosed reports and cover 24 core vulnerability classes.
  • It adds attack matrices for enterprise identity and infrastructure, plus Burp MCP integration.
  • The README says it was used on authorized red-team engagements and public training platforms.
topicany subjectsourcesReddit, X, HNscorevotes, likesbriefAI judge
+696 that day★ 69.1KPython

Agent skill that researches any topic across many sources

An AI agent skill that searches a topic across Reddit, X, YouTube, HN, Polymarket, and the web, then synthesizes a summary. Results are scored by upvotes, likes, and real money, not editors.

  • Searches run in parallel, then an AI agent judge synthesizes one brief.
  • Reddit, HN, Polymarket, and GitHub work with zero config.
  • A setup wizard unlocks X, YouTube, TikTok, arXiv, and more in about 30 seconds.
trynpx skills add mvanhorn/last30days-skill -g
git diffchanged filesagent + LLMreads, searchescommentsline-levelrulesetNPE, XSS, SQL
+763 that day★ 53.8KGo

AI code review CLI from Alibaba using an LLM agent

A code review CLI that reads Git diffs and sends changed files to an LLM through an agent with tool use. It produces structured review comments with line-level precision.

  • The agent can read full files, search the codebase, and inspect other changed files.
  • ocr scan reviews whole files, for auditing codebases with no meaningful diff.
  • The README benchmark claims higher precision and F1 than a general agent, at about 1/9 of the tokens.
tryocr scan
chatMain BotsandboxDocker + repospull requestreviewSOPhuman gates
+591 that day★ 3.3KGo

Self-hosted workflows for coding agents with approval gates

A self-hosted workspace where teams ship software with coding agents. Main Bot opens a Docker sandbox, checks out repositories, delegates to project roles, and opens the PR.

  • Workflows that prove themselves can be saved as SOPs: Agent, Script, Human-gate, and sub-SOP nodes.
  • It runs CodeBuddy, Codex, Claude Code, OpenCode, Cursor CLI, and others.
  • It runs on your own infrastructure under the MIT license.
Web SDKeditor, calcServer SDKcollab, filesAI SDKtools, reviewyour appembeds editoragentuses content
+1.8K that day★ 28.9KTypeScript

Office SDK for embedding Sheets and Docs in your product

Univer is an Office SDK for building Office applications inside your own product, with editing, calculation, collaboration, and file processing for people and agents.

  • The SDK has three parts: Web SDK, Server SDK for collaboration and import/export, and AI SDK.
  • The family covers spreadsheets, documents, presentations, relational tables, canvases, and PDFs.
  • This repo holds the Web SDK core architecture under Apache-2.0.
youadminArcanev3 branchDockermanagementdocsgetarcane.app
+618 that day★ 9.4KGo

Modern Docker management designed for everyone

Arcane is a Docker management app. The main branch is now v3, while the last 2.x release lives on the release/2.x.x branch.

  • Feature requests moved to Discussions, and issues are only for bug reports.
  • Setup, configuration, and development guides are on the official docs site.
  • Translations are handled through Crowdin.
agentCodex, Claudeskillsdesign, buildStitch MCPservermarketplace3 plugins
+350 that day★ 10.1KTypeScript

Agent Skills library for the Google Stitch MCP server

A library of Agent Skills that work with the Stitch MCP server and follow the open Agent Skills standard. Compatible with Codex, Antigravity, Gemini CLI, Claude Code, and Cursor.

  • Three plugins are provided: stitch-design, stitch-build, and stitch-utilities.
  • Install through the plugin marketplace in Codex, Claude Code, and Cursor.
  • OpenCode is supported through manual install.
trycodex plugin marketplace add google-labs-code/stitch-skills --ref main
plugin.jsonmanifestpluginown directoryCursormarketplaceexamplesteaching, ralph
+201 that day★ 16.9KTypeScript

Cursor plugin specification and official plugins

This repo holds the Cursor plugin specification and official plugins for popular developer tools, frameworks, and SaaS products. Each plugin is its own directory.

  • Each plugin has its own .cursor-plugin/plugin.json manifest.
  • Examples include teaching, continual-learning, ralph-loop, and create-plugin.
  • Other plugins cover branch review, repo compatibility scans, and CLI patterns for agents.
ideaagent asksspecshort chunksplanclear stepssubagentsTDD per taskreviewthen continue
+1.9K that day★ 319KShell

Agentic skills framework and software development methodology

Superpowers is a complete software development methodology for coding agents, built on composable skills. The skills trigger automatically, with nothing special to do.

  • The agent asks what you are trying to do first, then shows the spec in short chunks.
  • After sign-off it writes an implementation plan that stresses red/green TDD, YAGNI, and DRY.
  • A subagent-driven development process then works through each task, inspecting and reviewing the work.
agentany frameworkevent streamtyped eventsweb apprendermobilerenderterminalor Slack
+1.4K that day★ 20KTypeScript

Open protocol connecting AI agents to any app

AG-UI is an open, event-based protocol that standardizes how AI agents connect to user-facing apps. Version 1.0 is out with a frozen JSON Schema.

  • The agent emits typed events over one bi-directional stream, and any client can render them.
  • The TypeScript, Python, and .NET SDKs are generated from the same schema.
  • 0.x agents and 1.0 clients work with each other.
trynpx create-ag-ui-app@latest
agentplan, diff, htmlreviewlocal browserfeedbackone clickagentacts on it
+691 that day★ 13.1KTypeScript

Local review surface for coding agent plans and diffs

Plannotator is a local, browser-based review surface for AI coding agents. Plans, specs, HTML, and diffs can be annotated, and the feedback goes back to the agent.

  • It plugs into agents through plugins, hooks, and commands, including Claude Code, Codex, Gemini CLI, and OpenCode.
  • Code review covers local changes or remote PRs, with Git, GitButler, or Jujutsu.
  • Your comments go back to the agent with one click.
componentsdeclaredcompilerbuild stepJavaScriptefficientDOMlive updates
+477 that day★ 89.6KJavaScript

Compiler that turns components into efficient JavaScript

Svelte is a way to build web applications: a compiler that turns declarative components into efficient JavaScript that surgically updates the DOM.

  • The project is MIT-licensed, volunteer-driven, and accepts support through Open Collective.
  • A roadmap and contributing guide are available in the repo and on svelte.dev.
  • Donated funds cover costs such as hosting.
skillsinstinctsmemorysecurityECCharnessClaude CodeCodexCursorOpencode
+883 that day★ 306KJavaScript

Agent harness system: skills, memory, security, research first

ECC is a harness system for coding agents. It plans before it builds, verifies changes with tests, and reviews its own work from a fresh context.

  • It covers skills, instincts, memory, security, and research-first development.
  • It targets Claude Code, Codex, Opencode, Cursor, and more.
  • Install only from official channels: the GitHub repo, the npm packages, and the ecc@ecc plugin.
tokenizerfrom scratchpretrain64M modelSFT1 epochRLHF / RLAIFDPO, PPO, GRPO
+192 that day★ 66KPython

Train a 64M-parameter LLM from scratch in 2 hours

MiniMind trains a small language model of about 64M parameters from scratch. The core code is native PyTorch, with no high-level abstractions from third-party libraries.

  • The 2-hour claim is one SFT epoch on a single NVIDIA 3090.
  • It covers pretraining, SFT, LoRA, DPO, PPO, GRPO, and model distillation.
  • Apache 2.0 license, and it doubles as an intro tutorial to LLMs.
ideas + logsraw materialoutlineagentliteraturesearch, citesectionswriterefineplotsLaTeXmanuscript
+1.1K that day★ 3.1KPython

Multi-agent framework for writing AI research papers

PaperOrchestra turns ideas and experimental logs into submission-ready LaTeX manuscripts. Specialized agents handle outline, literature review, writing, refinement, and plotting.

  • The agent pipeline: Outline, Literature Review, Section Writing, Content Refinement, Plotting.
  • It searches, prioritizes, and integrates relevant literature into the manuscript.
  • LaTeX output follows the target conference template.