Free tool
Is this skill any good? Is it safe?
Paste a SKILL.md or drop a skill folder. You get a 0–100 score across four checks, every finding with its line, and the tokens it costs your agent. Works for Claude Code, Cursor, Codex and Gemini CLI skills.
name and description rules, YAML that parses, only the fields claude.ai and the Skills API accept, files that resolve
a description that triggers, body under 5,000 tokens, detail moved to files read on demand
ordered workflow, a concrete example, a defined output, no vague filler
prompt injection, concealment, credential theft, remote code, fake prerequisites, hidden Unicode, persistence
The same scanner runs locally. Add it to CI and a skill with a critical finding, or below your bar, fails the build before anyone installs it.
npx proskills-cli scan .claude/skills .agents/skills --min 80
Open source (MIT) on GitHub, with 11 free skills.
Scoring follows the open Agent Skills specification (agentskills.io) and published research on malicious skills. A clean scan is not a guarantee: static scanners can be evaded. Every ProSkills skill is also read line by line by a person before it ships.