Python
483 repositories, by attention — Page 14
🦀 Jumper — an crab robot.
Find the ghost tokens. Fix them. Survive compaction. Avoid context quality decay.
Security Scanner for Agent Skills
15 motion design styles, each made by Claude Opus 5.5 writing code from one prompt: watch the films, copy the prompts, read the full source.
Skill pack automates the full PR lifecycle for coding agents — A set of reusable agent skills for Claude Code and Codex that automate end-to-end PR workflows: proving, reviewing, filing, babysitting through CI, and merging changes, plus multi-model review loops. Built by t3dotgg from daily use, it suits developers running autonomous coding agents who want less manual PR babysitting.
The agent skills I use every day. Works with Claude Code and Codex.
Listen to different sounds
Where VS Code's bundled Python language server lives in public — This repo is the issue tracker and docs hub for Pylance, the closed-source language server that powers Python support in VS Code by default. It's useful for anyone debugging type-checking, completion, or diagnostics issues, and for those wanting to improve the underlying engine via Pyright, its open-source core.
Documentation and issues for Pylance
Install and manage GE-Proton, Luxtorpeda & more for Steam and Wine-GE & more for Lutris with this graphical user interface.
A voice assistant that lets you swap Gemini for a local or offline model — Mark LVI is a cross-platform voice AI built on Gemini Live, now extended with a pluggable engine layer for speech, LLM and TTS. You can route it to local models (Ollama, llama.cpp, vLLM) or other APIs (OpenAI, Groq, Mistral) without touching the voice pipeline, and the provider code loads at zero cost if unused.
Newest, latest and most advanced model that is able to control computers and even more. We're improving that model day by day, so follow up to know and see new models.
Daily radar for new AI terms: names of new models, products and repos, caught on X, Hugging Face, GitHub and sitemaps before keyword tools list them.
The low-level engine quietly powering boto3 and the AWS CLI — Botocore is the foundational, low-level interface to AWS services that underlies both boto3 and the AWS CLI. It's maintained directly by AWS, making it the base layer developers rely on indirectly whenever they use Python to talk to AWS. Recent notices track Python version support, including the end of Python 3.9 support in April 2026.
The low-level, core functionality of boto3 and the AWS CLI.
LinkedIn enumeration tool to extract valid employee names from an organization through search engine scraping
A 3D steering lab for every MLP neuron in Qwen3.5-4B — This tool sweeps every MLP neuron in Qwen3.5-4B up and down across prompts, keeps every rollout, and renders the results as a navigable 3D map. It adds live steered chat, logit/steer lenses, UMAP embeddings, and an MCP-connected Claude agent that can inspect neurons and drive the explorer itself. Useful for interpretability researchers probing concrete neuron behavior rather than abstract activations.
Steering-first neuron explorer for Qwen3.5-4B: push every MLP neuron, keep every rollout, browse the results
SDK for Reachy Mini, an open-source robot with an app store — Reachy Mini is an open-source expressive robot aimed at hackers and AI builders, supporting wireless, USB-connected Lite, and MuJoCo simulation setups with no hardware needed. This SDK also plugs into an app store built on Hugging Face Spaces, offering community apps like LLM conversation, radio, and hand tracking.
Reachy Mini's SDK
A practical hardware and software setup for a compact RL-powered biped robot.
Agent Skills that teach coding agents idiomatic Rust — Three Agent Skills plus a tested clippy config teach Claude Code, Codex, Copilot, Cursor, or Gemini CLI to write, review, and port idiomatic Rust. Guidance is distilled from five major style sources (Google, Rust API Guidelines, Effective Rust, Rust Design Patterns, Microsoft), with the skill picking a side when they disagree.
Agent Skills that teach coding agents to write, review, and port idiomatic Rust, plus a tested clippy config. Works with Claude Code, Codex, Copilot, Cursor, and Gemini CLI.
Full open-source build files for a buildable humanoid robot — Asimov 1 is a 1.2m, 35kg biped with 25 actuated degrees of freedom, and this repo publishes the mechanical and electrical CAD, simulation model, and onboard software needed to build, simulate, and customize it. It's aimed at hobbyists and researchers, with a $20,000 DIY kit or a full BOM for self-sourcing parts.
v1 of Asimov, an open-source humanoid robot
A Claude Code skill that chases citations back to primary sources — imozuru is a Claude Code skill that, instead of summarizing top search results, maps a field's terms, key people, and venues, then follows citations, links, and authors until it finds and saves primary sources. v2.0 adds human-readable detail reports and a dashboard, plus a watch mode that tracks new posts via RSS. It's for users who want to verify the evidence behind an AI's answers, not just read a summary.
知らない分野を、原文にたどりつくまで芋づる式に掘る Claude Code のスキル
Metadata harvester
Python SDK for local and cloud background removal (pip install withoutbg)
NVIDIA's Modular Closed-Loop Simulator for End-to-End AV Research — AlpaSim is an open-source, Python-based closed-loop simulator from NVlabs for testing end-to-end autonomous driving policies. It pairs pluggable sensor rendering (NuRec, plus video-model rendering via OmniDreams/FlashDreams) with a microservices architecture, letting researchers swap components and scale across nodes for algorithm validation and benchmarking.
AlpaSim is an open-source autonomous vehicle simulation platform designed for development and testing of end-to-end AV policies
Evidence-first genealogy skill for AI agents (Codex, Claude Code): church books, civil, military, deportation and repression records across the former USSR. Replies in your language.
Your iPhone helps your Mac run a 27B model: faster prompt reading and more context over a USB-C cable
Five Coordinated Qwen3.8 Instances Rival a Frontier Model on Hard Coding Benchmarks — This repo introduces ledger-based zero-shot self-orchestration: fresh instances of one model decompose problems and coordinate via a shared filesystem holding a plan, notes and current solution. On 100 hard LiveCodeBench problems, orchestrated Qwen3.8 models approach or slightly exceed a frontier model's accuracy at a fraction of the cost, including on self-hostable weights. Useful for researchers exploring inference-time scaling instead of bigger models.
GVS5H: Five Qwen3.8-27B Models Match Claude Fable 5 on LiveCodeBench Hard | Fable 5 Level Coding for a Fifth the Price - or on a Single GPU
A single-file Python PoC that replays a Bitcoin CTF exploit chain — This repo publishes a standard-library Python script reproducing the exact request sequence that solved the Steal My Bitcoin CTF, including PHP heap-leak and code-leak OOM stages against a Dockerized Alpine/PHP 8.4.26/nginx lab. It is useful for anyone studying real-world PHP memory-disclosure exploitation or reproducing CTF writeups end to end rather than reading theory.
A public release of the Steal My Bitcoin challenge
AI assistant and MCP server for Blender: visual chat, scene tools, multi-agent workflows. Works with OpenAI, ChatGPT, Claude, Gemini, Ollama, OpenCode. Blender AI addon and plugin, copilot for 3D, bpy Python scripting, MCP integration.
Free Claude skill: make Claude build award-level, cinematic landing pages. By MotionSites.
Source-linked Claude 5.5 video workflows: Opus and Sonnet demos, separately labeled next-Fable testing reports, prompts, tools and production evidence.
NVIDIA's World-Simulation Model Steps Aside for Cosmos 3 — Built on Cosmos-Predict2.5, this model generates world simulations from multiple spatial control inputs for physical AI research. The repo is now in limited maintenance, as NVIDIA directs development toward the unified Cosmos 3 platform, which merges reasoning, prediction, and transfer into one architecture.
Cosmos-Transfer2.5, built on top of Cosmos-Predict2.5, produces high-quality world simulations conditioned on multiple spatial control inputs.
Official Hi3D v1.0: high-resolution 3D geometry refinement built on Sparc3D. Inference code, model weights, and documentation.