**Added:**
- **NVIDIA / SkillSpector** — Security scanner for AI agent skills to detect vulnerabilities, malicious patterns, and security risks. (2,241 total stars, 308 stars today)
- **hexo-ai / sia** — A self-improving AI framework that autonomously boosts the performance of any AI system on a benchmark task. (996 total stars, 177 stars today)
- **francescopace / espectre** — Motion detection system using Wi-Fi spectre analysis (CSI) with Home Assistant integration. (8,516 total stars, 98 stars today)
- **karpathy / autoresearch** — AI agents that automatically run research on single-GPU nanochat training. (86,103 total stars, 201 stars today)
- **onyx-dot-app / onyx** — Open-source AI platform with advanced chat features compatible with every LLM. (30,236 total stars, 30 stars today)
- **LLMQuant / quant-mind** — Intelligent knowledge extraction and retrieval framework for quantitative finance. (1,036 total stars, 336 stars today)
- **modelscope / FunASR** — Industrial-grade speech recognition toolkit supporting 50+ languages, speaker diarization, emotion detection, streaming, and an OpenAI-compatible API. (17,792 total stars, 96 stars today)
- **anthropics / claude-agent-sdk-python** — Python SDK for Claude agents (7,266 total stars, 12 stars today)
- **mlflow / mlflow** — Open-source AI engineering platform for agents, LLMs, and ML models with debugging, evaluation, monitoring, and cost control. (26,423 total stars, 22 stars today)
- **zulip / zulip** — Open-source team chat server and web application. (25,341 total stars, 6 stars today)
- **anthropics / claude-plugins-official** — Official directory of high-quality Claude Code Plugins managed by Anthropic. (29,875 total stars, 83 stars today)

**Changed:**
- **mvanhorn / last30days-skill** — total stars increased from 39,444 to 39,476; forks from 3,173 to 3,176.
- **soxoj / maigret** — total stars increased from 32,296 to 32,322; forks from 2,372 to 2,374.
- **maziyarpanahi / openmed** — total stars increased from 2,473 to 2,492; forks from 252 to 253.
- **alistaitsacle / free-llm-api-keys** — total stars increased from 2,188 to 2,191; forks unchanged at 223.
- **BerriAI / litellm** — total stars increased from 50,028 to 50,032; forks unchanged at 8,785.
- **Sumanth077 / Hands-On-AI-Engineering** — total stars increased from 2,017 to 2,021; forks unchanged at 563.
- **FareedKhan-dev / train-llm-from-scratch** — total stars increased from 5,473 to 5,493; forks from 727 to 730.

**Removed:**
- **harry0703 / MoneyPrinterTurbo** — Uses AI large models to generate high-definition short videos with one click. (85,610 total stars, 1,389 stars today)
- **luongnv89 / claude-howto** — Visual, example-driven guide to Claude Code from basics to advanced agents with copy-paste templates. (36,778 total stars, 211 stars today)
- **roboflow / supervision** — Reusable computer vision tools. (43,757 total stars, 695 stars today)
- **google / skills** — Agent skills for Google products and technologies. (13,482 total stars, 211 stars today)
- **Andyyyy64 / whichllm** — Finds the local LLM that performs best on your hardware using real, recency-aware benchmarks. (4,484 total stars, 479 stars today)
- **davila7 / claude-code-templates** — CLI tool for configuring and monitoring Claude Code. (27,924 total stars, 35 stars today)
- **RyanCodrai / turbovec** — Vector index built on TurboQuant with Python bindings. (10,931 total stars, 770 stars today)
- **ml-explore / mlx-examples** — Examples in the MLX framework. (8,715 total stars, 13 stars today)
- **vllm-project / vllm-omni** — Framework for efficient inference with omni-modality models. (5,090 total stars, 24 stars today)
- **comet-ml / opik** — Debug, evaluate, and monitor LLM apps, RAG systems, and agentic workflows. (19,573 total stars, 44 stars today)