Trending
skills & tools · github stars- Vincentwei1021/anything2explainer
Topic in, narrated explainer video out. A Claude Code / Codex skill that turns any topic into a black-canvas motion-graphics explainer video with TTS voiceover, subtitles and a chapter progress bar. C
- Nanako0129/sepia
De-AI writing skill for any Agent Skills-compatible agent (77+ via the Skills CLI), with native plugins for Claude Code, Codex, Grok Build, and Antigravity. Narrative-architecture repair for fiction,
- mizzlelover/gongwen-gbt9704-skill
跨平台中文公文排版 Skill，生成参照 GB/T 9704-2012 的可编辑 DOCX
- dmmulroy/anti-slop
Opinionated Oxlint rules for rejecting low-evidence TypeScript and JavaScript patterns
- nateherkai/scroll-craft
An agent skill for building premium, immersive, scroll-driven websites. Works with Codex, Claude Code, and other coding agents. Also available as a Claude Code plugin.
- 2akouwu/reverify
Stop your AI from making things up — it proposes, deterministic tools decide, every claim checked against ground truth with evidence. Grounded facts and context survive resets. Reverse engineering is
Tools & Packages
- ripwire: ripgrep of AI context (CLI+MCP) giving coding agents a map of any repo
ripwire is a CLI+MCP tool that lets coding agents quickly get a contextual map of a repo, combining ripgrep semantics with AI-friendly context injection.
- Claude Code skills for advanced context engineering techniques and patterns
A curated kit of Claude Code skills for advanced context engineering, providing reusable patterns to improve agent reasoning and reduce drift.
- Coop – Isolated VM Environments for Running Claude Code and Codex
Coop provides isolated VM environments for running Claude Code and Codex, addressing security and reproducibility for agent-based workflows.
- The VMs Powering Mobile Agents (Instinct, Claude Code)
An analysis of VM platforms powering mobile agents like Instinct and Claude Code, crucial for understanding how agent environments are sandboxed and orchestrated.
- Self-hosted company OS, Claude Code and Codex agents in departments
A self-hosted multi-tenant company OS integrating Claude Code and Codex agents into department workflows, enabling collaborative agent use with local or cloud models.
Patterns & Techniques
- Portal by Spotify cut my Claude Code token usage by 90%
Spotify's 'Portal' technique cuts Claude Code token usage by 90% via smart context caching—a must-read pattern for anyone building cost-efficient agent integrations.
- Video compressor
Demonstrates using Claude Code with Fable 5.1 to build a WebAssembly-based FFMPEG video compressor tool from a prompt example. Shows how to rapidly prototype and deploy custom tooling via coding agents.
- Using Blender with coding agents on macOS
Shows a practical pattern for integrating Blender with coding agents on macOS via a simple prompt. Highlights how agents can drive creative 3D tooling without manual setup.
Discussion
- Building Codex with Tibo Sottiaux
An inside look at how Codex (OpenAI's coding agent) was built, offering insights into its architecture and the vision for AI-driven software development.
- Qwen3.8-27b is the first Local model im able to blindly trust
User reports Qwen3.8-27B performing continuous agentic work for 8+ hours without failures using a custom agentic loop in a harness, highlighting trust in local models for coding tasks.
Research
- Research acceleration: The view inside OpenAI
Describes OpenAI's internal use of coding agents (agentic engineering) accelerating their own research, calling 2026 the year it took off. Must-read for understanding how top labs leverage agentic workflows to accelerate AI R&D.
- When Passing Tests Hides Vulnerabilities: An Empirical Study of Silent Failures in Agentic Systems
Empirical study finds that patches passing tests in LLM-based agentic code repair can still introduce security vulnerabilities; systematically categorizes silent failures.
- What a Random Draw from the MCP Registry Contains, and What Tool-Use Benchmarks Contain Instead
Probability sample from MCP registry shows only 48.8% of servers complete an invocation, revealing selection bias in existing tool-use benchmarks and studies.
- Engineering Reliable Commit Gates for Agentic AI: Cost-Aware Verification Portfolios under Common-Mode Data Failures
Proposes VP-CONTROL, a runtime assurance design and benchmark for cost-aware commit gates in agentic systems, with 2,880 scenarios across fault regimes.