wtf is skills?

StarSling Skills is a repository that currently ships one tool, called ci-speedup. This tool audits GitHub Actions workflows, which are the automated checks that run when a developer submits code changes. The goal is to figure out why those checks are slow and produce a report explaining the actual causes of the delay. The tool measures slowness along two separate axes. The first is developer wait time, meaning the slowest checks a developer must sit through before their code can be merged. The second is runner minutes, which translates to the cloud computing bill. These two numbers are kept apart because a change might reduce the bill while adding wait time, so combining them into a single score would hide that tradeoff. The tool works from real run history pulled through the GitHub API, not from guesses or estimates. The audit draws from a catalog of over 70 optimization patterns across 14 categories. These include missing caches, redundant setup steps, sleep-based readiness checks, and unsharded test jobs. An important distinction is that the tool diagnoses problems but does not prescribe fixes. It cannot tell whether something that looks wasteful was done on purpose. Instead of generating a patch, each finding includes a prompt you can paste into your own coding assistant. That assistant then reads the actual logs and file history to propose a safe change. Installation uses a command line tool called npx. You select your coding agent, such as Claude Code, Codex, or Cursor, and the skill installs into that environment. Running it requires the GitHub command line tool to be authenticated, plus Python 3.9 or newer and a library called PyYAML. The tool samples your recent CI runs, and the process takes roughly a minute on a mid-size repository. It sends nothing to StarSling or any third party. The full report is optional and can be saved as a markdown file in your working directory.