vibe-security-skill

Why we included this project

Apps assembled largely by AI coding assistants tend to carry the same recurring security flaws: keys hardcoded into source, row-level security left off, prices trusted straight from the browser, tokens parked in localStorage. This skill packages those failure patterns so your coding agent can audit its own work before anything ships. It follows the Agent Skills format and installs into Claude Code, OpenAI Codex, and other compatible agents with a single command. The rules live in separate reference files keyed to your stack, so it audits Supabase RLS policies when it sees Supabase, payment flows when it sees Stripe, and bundled secrets when it sees React Native, skipping checks that don't apply. If you trust an AI to move fast but still want a cautious security pass before merging, this fills a real gap without adding another tool to keep up with.