skills

Capture and attach UI evidence for pull requests. Use when Codex opens, updates, prepares, or reviews a PR with user-visible UI changes and needs screenshots, video, or Playwright trace evidence; when a user asks to add PR screenshots/video; or when a PR body lacks visual evidence for UI work.

Use this skill when the user wants to set up or run iterative experiments to improve a single metric over time, including eval-score optimization, benchmark tuning, prompt iteration, hill-climbing, keep-or-revert workflows, or evolutionary testing. This skill interviews the user, helps bootstrap a cheap benchmark when none exists, scaffolds the experiment workspace, and runs one-mutation keep/discard loops.

This skill should be used when the user wants to set up an automated Claude agent loop that works through a list of tasks iteratively. It scaffolds a ralph loop — a bash-driven harness that runs Claude repeatedly, picking up the next work item each iteration until all items are complete. Triggers on "set up a ralph loop", "create an agent loop", "ralph", "loop through tasks", "automate this work", or when the user has a list of items they want Claude to work through autonomously.

This skill should be used when the user asks to "address PR comments", "handle PR feedback", "review PR comments", "go through PR comments", "resolve PR comments", "process PR review", "fix PR comments", "respond to review feedback", "address code review comments", or mentions dealing with pull request comments, code review feedback, or reviewer suggestions.