Traction scoreGitHub stars can be faked, so popularity alone can be misleading. Traction Score looks for broader signs of recent attention, adoption, and active maintenance.
This skill enables the capture of side-by-side screenshots of web pages or specific elements to document visual changes. It is designed for developers who need to provide clear evidence of UI updates in pull requests or perform visual comparisons between different URLs and image paths.
WHO IT'S FOR
AI agent / automation builders
implementing evidence-driven testing for AI agents
Frontend / UI engineers
documenting visual changes with before-after screenshots
Technical leads / architects
refactoring operational logic into service layers
Platform / DevEx teams
standardizing AI coding workflows and review loops
Compatible AgentsThe repository documents support for these agents. The skills may also work with other agents that can load SKILL.md files, but they may need some setup or small changes.
Follow the documented setup, then try a first task.
Install the skills collection for Claude Code
Installs: Complete michaelshimeles/skills collection; the listed skill paths are a bounded selection of skills contained in the bundle. · Claude Code
Before you start
npx must already be available.
For before-and-after comparisons: the @vercel/before-and-after CLI, available through npx or the documented global package installation.
For recorded evidence: Python 3, ffmpeg and ffprobe built with libx264 and the ass filter; an agent-controlled GUI and authenticated browser session. Linux uses X11/x11grab or Wayland/wf-recorder; macOS needs Screen Recording permission; Windows uses gdigrab. Alternatively, the headless path needs a running app and a scriptable browser. Posting evidence requires gh or an equivalent tool.
For greploop or greploop-apps: git plus authenticated gh or glab and Greptile installed on the repository; the Perforce path requires authenticated p4.
Run this terminal command to add the michaelshimeles/skills collection. Claude Code will pick up matching skills automatically.
Review duplicated logic for service-layer extraction
Uses code-structure
Paths to two or more action or workflow files that may duplicate operational logic
Any repository-specific architecture conventions
Use the code-structure skill to review the supplied action or workflow files. Identify one operational block repeated across at least two callers, recommend what should move into a shared service, and state what domain policy should remain in each action. Return a small extraction plan with explicit service inputs and structured outputs; do not edit the files.
Use the unslop skill to rewrite the supplied draft for a human reader. Preserve its meaning and intended tone, remove obvious AI patterns, vary the rhythm, and make vague or promotional wording more specific. Return only the revised draft; do not post or send it.
Claude
Codex
Cursor