8 Must‑See Trending GitHub Open‑Source Projects This Week

1. scientific‑agent‑skills

An out‑of‑the‑box AI skill bundle that covers research, scientific computing, engineering, data analysis, finance, and writing. It provides a ready‑to‑use set of Agent skills, enabling AI to follow a structured workflow rather than producing erratic outputs. The project recently broke 25 000 Stars and continues to climb.

Open‑source URL: https://github.com/K-Dense-AI/scientific-agent-skills

2. academic‑research‑skills

A specialization of the previous bundle focused on the end‑to‑end academic paper workflow: literature search, drafting, reviewing, revising, and finalizing. The pipeline mirrors the real‑world writing process, requiring occasional human intervention. It gained over 10 000 Stars in a single week, approaching 20 000 Stars.

Open‑source URL: https://github.com/Imbad0202/academic-research-skills

3. Understand‑Anything

A code‑base visualizer that turns any repository into an interactive knowledge graph, supporting search, Q&A, and visualization. It works with multiple AI assistants (Claude Code, Codex, Cursor, OpenCode) so that the model instantly understands the project structure without re‑parsing each time. The repo is close to 20 000 Stars.

Open‑source URL: https://github.com/Lum1104/Understand-Anything

4. codegraph

Indexes an entire codebase into a graph before feeding it to an LLM, allowing the AI to grasp the project instantly. Supports Claude Code, Codex, Cursor, and other mainstream tools. The approach reduces token consumption and improves edit accuracy, especially for large projects.

Open‑source URL: https://github.com/colbymchenry/codegraph

5. oh‑my‑pi

A terminal‑based AI coding assistant that emphasizes precise code modifications. It now has over 6 000 Stars and includes a Hashline editing system that uses content‑hash anchors to avoid token waste, cutting token consumption by about 61 % and improving edit reliability.

Open‑source URL: https://github.com/can1357/oh-my-pi

6. 12‑factor‑agents

Applies the classic 12‑factor app principles to AI agents, defining a deterministic, stateless workflow for tool invocation, prompt management, context handling, and error handling. The repository has surpassed 21 000 Stars and provides three hands‑on workshops plus scaffolding to bootstrap compliant projects.

Open‑source URL: https://github.com/humanlayer/12-factor-agents

7. ai‑engineering‑from‑scratch

A comprehensive learning path (428 lessons, ~320 hours) that walks from linear algebra to autonomous multi‑agent systems. Each lesson includes problem definition, conceptual explanation, implementation in Python, then a PyTorch or scikit‑learn version, finally delivering a deployable AI artifact. The repo holds about 13 000 Stars.

Open‑source URL: https://github.com/rohitg00/ai-engineering-from-scratch

8. Supertonic

An offline, on‑device text‑to‑speech system (~99 M parameters) that runs in real‑time on CPU via ONNX Runtime. It supports 31 languages, offers expression‑tag control for emotional speech, and provides SDKs for 11 platforms (C++, Node.js, Python, Rust, etc.). The project has over 6 000 Stars.

Open‑source URL: https://github.com/supertone-inc/supertonic

9. ViMax

A multi‑agent video production suite from HKU Data Science Lab that decomposes video creation into AI roles (director, screenwriter, producer, generator). It supports Idea‑to‑Video, Script‑to‑Video, and Novel‑to‑Video pipelines, and includes an AutoCameo feature that inserts a user’s photo as a character. The repo has accumulated over 1 000 Stars.

Open‑source URL: https://github.com/HKUDS/ViMax