AI Web Design & Frontend Skills Catalog · Question revision 1 · Which ready-made AI agent skills are most useful for web design and frontend development, and when should you use each one?

12 October 2026 · Agent Browser source recheck

Agent Browser's SKILL.md is a discovery stub.

The v0.38.1 file linked from our catalogue is a discovery stub. It tells the agent to load the full workflow from the installed CLI with agent-browser skills get core. If you read only the linked file, you have not read the instructions the author expects the agent to use.

What to check before a frontend trial

1. Keep the tagged stub as the source pin, and inspect the tagged core file for the actual navigation, snapshot, and screenshot workflow.
2. In your own environment, record the installed Agent Browser package version and the output of agent-browser skills get core. A repository tag does not prove what the agent received at runtime.
3. Name the route, action, expected visible state, viewport, and session before judging a capture. For exploratory bug hunting, the stub points to agent-browser skills get dogfood as a separate workflow.

Evidence limit: We re-read the tagged files. We did not install the CLI, check Codex loading, run a browser, capture a screen, or compare a result with a no-skill brief. The core file describes a workflow; it does not prove that the workflow catches a defect or improves a design.

Catalogue correction

The Agent Browser card now links both the discovery file and the full tagged core source. The listing remains a conditional fit for a runnable UI state that needs browser evidence, not a design or accessibility verdict. The next useful test is a fixed frontend flow with the package version and loaded instructions retained alongside the browser artifacts.