GitHub AI Daily Ranking: New Agent Skills 'rust-anything' and 'inference-optimisation-skill' Surge to 2nd and 3rd Places
In the GitHub AI daily rankings for November 18, 2026, 'rust-anything' for progressive Rust rewriting and 'inference-optimisation-skill' for local LLM speed tun

Image source: https://synon.co.jp/article/news/2026-11-18-github-ai-daily
In the GitHub daily trending rankings for AI-related repositories published on November 18, 2026, newly released agent skills created just days prior have surged to the top of the leaderboard, demonstrating a rapid shift in developer tooling.
Image Credit: Synon / GitHub Search API aggregation
While the Flutter-based remote control application dsh-Go maintained the overall top spot, two open-source projects registered just two days prior on November 16 captured the 2nd and 3rd positions respectively. Beyond merely assisting with initial code generation, these specialized skills aim to restructure existing software architectures and optimize large language model performance.
Progressive Rust Rewriting with 'rust-anything'
Claiming the 2nd spot is erkamyaman/rust-anything (gaining 822 stars), an agent skill designed to help rewrite existing legacy applications into Rust—a systems programming language renowned for memory safety and high performance—gradually and incrementally.
- How It Works: It performs refactoring by cross-referencing execution logs and behavior records of the target application, which may be originally written in TypeScript, Python, C, Go, Ruby, or Java (such as CLI tools, HTTP servers, or libraries).
- Core Objective: It minimizes human errors that typically occur during manual porting and allows an agent to systematically verify each step of code migration.
The repository explicitly lists ecosystem integrations under its topics, including Claude Code (Anthropic's official CLI tool for interacting with AI in terminal environments), Codex, and the Model Context Protocol (MCP), a standard communication protocol enabling AI models to safely access external data and tools.
Hill-Climbing Local LLM Optimization with 'inference-optimisation-skill'
Ranking 3rd is sybil-solutions/inference-optimisation-skill (gaining 790 stars), an agent skill focused on maximizing the processing speed of local large language models running on personal hardware or self-hosted servers.
- Optimization Approach: It employs a hill-climbing optimization algorithm, iteratively evaluating neighboring states to continuously improve performance.
- Target Areas: The skill addresses major inference bottlenecks, including decoding, speculative decoding, prompt prefilling, Key-Value (KV) cache handling, startup times, and Time To First Token (TTFT).
Developers have welcomed its hardware-agnostic design, which operates independently of specific chip architectures or quantization schemes.
Entering the Era of 'Agent Skills'
The November 18, 2026, GitHub AI daily ranking illustrates that developer interest is shifting from basic chatbot interfaces and foundation model weights toward specialized agent skills that directly target real-world engineering workflows.
Alongside the top contenders, repositories like boringmarketer/boring-marketing-skills (ranking 10th with 473 stars), which packages 31 marketing-related agent skills, highlight a broader trend. Developers and creators are moving past broad prompts, instead combining modular, task-specific agent skills to achieve immediate productivity gains in production environments.
Sources
- Synon: GitHub AI 関連リポジトリランキング(2026年11月18日)
- GitHub Repository: erkamyaman/rust-anything
- GitHub Repository: sybil-solutions/inference-optimisation-skill
- GitHub Repository: ccch1mneyyy/dsh-Go
- GitHub Repository: boringmarketer/boring-marketing-skills