A new Claude Code skill called "gold-digger" dropped on Hacker News yesterday, and it's got the kind of quiet confidence that makes you wonder why nobody thought of this sooner. Built by developer jaimeramiro-dev, the project lives at github.com/jaimeramiro-dev/gold-digger and does exactly what its name suggests: it digs through your system looking for dev tools—then reports back only when it strikes something useful.
How Claude Code Skills Work
For those unfamiliar with the ecosystem, Claude Code skills are modular extensions that hook into Anthropic's CLI assistant. They're essentially curated prompt templates with embedded logic that let you tailor Claude's behavior for specific workflows. Instead of writing a novel-length system prompt every time you want specialized help, you load a skill and get contextual awareness without the boilerplate overhead. Gold-digger takes this pattern and applies it to one of the most tedious parts of onboarding onto an unfamiliar codebase: figuring out what tooling is actually available on your machine.
The Silent Treatment Is the Feature
What makes gold-digger interesting isn't just that it finds tools—it's that it knows when not to bother you. Many dev environment detection scripts flood you with output about every binary in your PATH, regardless of whether you care. Gold-digger appears to take a more selective approach: if nothing relevant surfaces, you're never bothered with the information at all. This is exactly the kind of UX restraint that separates useful tooling from noise-generating wrapper scripts that claim to "help" but actually just add cognitive load.
Why This Matters for Infrastructure Teams
If you work in platform engineering or developer experience, you've probably written some variation of this detection logic yourself—for CI/CD pipelines, Docker image builds, onboarding scripts, or environment validation checks. Gold-digger gives Claude Code users a reusable primitive that handles the discovery layer. Rather than hardcoding which tools you're checking for across ten different scripts, you can lean on one skill and let Claude reason about what it found in context of whatever task you've asked it to tackle.
The Bottom Line
This is a small project with a low HN score (8 points), but it's solving a real problem that gets reinvented constantly. If you're already all-in on Claude Code for your workflow, gold-digger slots in cleanly without adding friction—and the "only speak when you have something useful" philosophy is worth copying into more tools.