This daily update tracks GitHub repositories gaining momentum by measuring the percentage increase in star counts over a rolling seven-day window. By prioritizing relative growth over absolute totals, we surface emerging tools like ReelMimic—which saw a 461% surge in interest—rather than just established industry giants. This week’s leaders highlight a shift toward specialized utility, from Phone Harness MCP’s mobile device connectivity to DYZ MedPaper’s 32 specialized skills for clinical research automation. The data reveals a concentrated focus on infrastructure and specialized workflows, exemplified by Agent Foundation’s scalable runtime and kyy-logs/code-review-gate-skill’s auditable review protocols.

#RepositoryStars GainedGain %Overview
1edenfunf/reelmimic1572461%An AI-driven video production tool that recreates the style, rhythm, and framing of a reference video to generate original 2D animations.
2phone-harness/mcp612370.9%This Model Context Protocol (MCP) connector enables AI agents to interact with real iPhones and Android emulators in the cloud via screen viewing, tapping, typing, and app management.
3kyy-logs/code-review-gate-skill333277.5%Auditable code review skill for Claude Code and Codex: four independent probes, evidence-backed findings, and a JSON-derived merge gate.
4duyunzhoudyz/dyz-med-paper-skills241241%A collection of 32 specialized Agent Skills for Claude Code and Codex designed to automate medical research workflows, from literature retrieval to manuscript revision.
5converge-ai-labs/agent-foundation662232.3%An open-source runtime for building, running, and managing scalable AI agents on private infrastructure with durable execution and observability.

ReelMimic (edenfunf/reelmimic )

ReelMimic is an automated video generation platform that transforms reference videos into original 2D animations using a multi-agent AI workflow. The system analyzes technical elements like shot length, BPM, and camera movement to create a storyboard and plan for user approval. It supports various rendering engines, including watercolor, pixel art, and anime cel, allowing users to generate custom content via Claude Code or Codex. This tool is designed for creators looking to replicate specific visual styles without reusing original assets or footage.

Phone Harness MCP (phone-harness/mcp )

Phone Harness MCP provides a bridge between AI agents and mobile operating systems through a standardized MCP interface. It allows users to connect clients like Claude, Cursor, and ChatGPT to dedicated cloud iPhones or Android emulators. Key capabilities include screen capturing, element recognition, coordinate-based tapping, and app launching. This tool is designed for developers and enterprises looking to automate mobile workflows or test AI agent interactions within real mobile environments.

kyy-logs/code-review-gate-skill (kyy-logs/code-review-gate-skill )

Auditable code review skill for Claude Code and Codex: four independent probes, evidence-backed findings, and a JSON-derived merge gate.

DYZ MedPaper (duyunzhoudyz/dyz-med-paper-skills )

DYZ MedPaper provides a suite of 32 specialized Agent Skills designed to assist clinicians and medical researchers through the entire publication lifecycle. The repository enables AI agents to perform specific tasks such as PubMed literature searching, statistical result interpretation, and drafting English manuscripts from Chinese research ideas. Users can integrate these skills into Claude Code or Codex to manage complex workflows, including responding to reviewer comments and organizing submission materials. It is specifically tailored to bridge the gap between clinical data collection and formal academic writing.

Agent Foundation (converge-ai-labs/agent-foundation )

Agent Foundation (a13n) provides a self-hosted runtime designed to manage complex AI agent workflows at scale. The platform features a web-based console for debugging tool calls and a suite of SDKs in Python, TypeScript, Go, and Rust for seamless application integration. Key capabilities include durable execution through checkpoints, support for long-running tasks that persist after client disconnection, and granular visibility into workspace usage and costs. It is built for developers and enterprises needing to deploy reliable, observable agents within their own secure infrastructure.