Case Radar: Stop Reading Menus, Get the Real Dishes

Where this fits

You just heard about a new tool—say, an AI coding framework or a design system—and you want to know what people are actually doing with it. Not the GitHub README, not a Medium recap. Real screenshots, real deployed demos, real source code. Case Radar turns "scan the ecosystem for real artifacts" into a repeatable agent workflow, outputting a browsable HTML case collection instead of another text dump.

Why agents benefit

• Artifact-first research: The skill explicitly hunts for "real things" (screenshots, source code, live demos) rather than repo homepages or documentation—a distinction most research agents miss.
• Multi-stage pipeline with checkpoints: scan → recon → capture → embed. Each stage is interruptible, so you can say "stop," "skip," or "go back" mid-flow.
• Recon heuristics: A dedicated reference file (recon-heuristics.md) encodes where genuine artifacts tend to live, so the agent isn't guessing.
• Real capture, not description: Uses curl for direct resources and agent-browser for screenshots—the output contains actual visual evidence, not paraphrased claims.
• HTML output, not chat noise: The final deliverable is a browsable HTML case set with embedded materials, suitable for sharing or archiving.

Practical scenarios

Evaluating a new AI framework: You're considering adopting a new agent framework. Instead of reading its docs, Case Radar scans Twitter, blogs, and GitHub for real projects built with it, captures screenshots of running apps, and gives you a visual sense of what's possible.

Design system adoption: Before committing to a component library, you want to see real products using it. The skill finds production sites, captures their rendered pages, and shows you the actual visual outcomes—not just API references.

Ecosystem sweep for product ideas: You have a vague sense that "people are doing interesting things with X." Case Radar gives you a concrete, visual map of the landscape—far more useful for brainstorming than a list of repo links.

Add it to your agent workflow

npx skills add yunshu0909/yunshu_skillshub --skill case-radar

Then trigger it naturally: "看看大家用 X 做了什么" or "scan the X ecosystem for real cases." The skill auto-checks that agent-browser is installed (one-line install if not) and proceeds through its pipeline. More details on the skill page.

Compare before adopting

Before adding it permanently, compare:

• Quality signals: Does the recon heuristic actually find real artifacts, or does it fall back to READMEs? Test with a topic you know well.
• Maintenance freshness: Check the GitHub repo's commit history. Stale heuristics mean stale results.
• Alternatives: For deep, goal-oriented research, long-research is a better fit. For writing or PRD work, use writing-assistant or prd-doc-writer. Case Radar is specifically for "show me real examples."
• Workflow fit: Do you actually need HTML output? If you just want a quick answer, a plain chat response is more appropriate.

Why it is worth tracking

Case Radar addresses a real gap in agent research: most agents summarize text, but few go out and capture visual artifacts. With 753 stars and an active author, the skill has community validation. The maintainer's design philosophy—"menus are worthless, dishes are the real thing"—is a useful corrective to the text-centric defaults of most LLM research flows. Track it if you regularly need ecosystem scans or visual evidence in your agent workflows; evaluate it when your research tasks move beyond "summarize" into "show me what people actually built."