Claude Code: Good skills, bad skills

Brian Scanlan on what Intercom learned running an internal marketplace of Claude Code skills, where 153 contributors, about 31% of R&D, wrote 267 skills in the first three months. At that volume the question stops being whether to write skills and becomes how to tell the useful ones from the ones quietly making things worse.

His good skills do one thing, wrap it in a real feedback loop, and lean on deterministic scripts rather than hoping the model improvises correctly. They come with maintained evals, disclose context progressively instead of dumping everything up front, carry safety guardrails, and have a named owner. The bad ones fail in recognisable ways: they produce the same errors every run, nobody checks their output, their evals are graded by something unqualified to grade them, they are too big for Claude to actually read, or they are hard to invoke so nobody does.

The maintenance point is the one that generalises past Claude Code:

Skills rot silently otherwise: one person fixes theirs, everyone else keeps running the broken copy.

Their skills are open source at intercom/2x-skills, organised into plugins for skill tooling, security, test, code review and PR work. Related notes from the same team: AI approving PRs and 2x engineering velocity.