Awesome Claude Skills (Composio)

A curated, opinionated list of Claude Skills from the Composio team. Think of it as the safe starter pack for Claude Code and Claude.ai users who want to extend their agent without walking into the minefield Snyk's ToxicSkills report just mapped out. The timing matters. In April 2026, Snyk scanned 3,984 publicly distributed Claude Skills and found that 36.82% contained security issues and 13.4% had critical-level flaws — prompt injections, hidden instructions, malicious shell scripts, the whole list. GitHub responded by shipping gh skill preview so you can inspect a skill before it runs on your machine. But inspection only helps if you know what to install in the first place. That is exactly what this repo solves. Composio curates by category: code review, testing, PDF handling, SEO, documentation, data processing, creative writing, and more. Each entry links to the skill's source and describes what it actually does in plain English. The maintainers vet for quality signals — clear provenance, sensible permission requests, no hidden prompt-injection payloads — before adding to the list. It is not a security audit, but it is a much higher bar than "trending on GitHub." If you just started using Claude Skills and don't want to be part of the 36% statistic, this is the list to bookmark first. Pair it with gh skill preview for inspection, immutable release pinning, and a dedicated sandbox user, and you have a sane baseline for skill adoption in 2026.