AI and agent security / FREE TOOL
LLM skill review
Inspect SKILL.md metadata and instructions for capability declarations, remote execution, sensitive data access, and trust-boundary review signals.
02 Your results
A closer look starts here.
Add your input and run the check. We will show the observations and explain what needs review.
A QUICK WALKTHROUGH
How to use it
- 01Paste a SKILL.md file or select it locally.
- 02Inspect metadata and line-level instruction review signals.
- 03Verify the publisher, referenced files and effective host permissions before installing or enabling the skill.
What this check can tell you
Local, static review of one Markdown skill file. Signals can describe legitimate documented tasks or quoted examples; they are not proof of malicious intent. Referenced files, downloaded scripts, model behavior and host permissions are not inspected. Maximum size: 200,000 characters.
The initial review processes files in this tab. If you choose the optional AI summary, the displayed findings are sent to Cloudflare AI; your raw file, result tables and decoded token claims are excluded.
Agent Skills format specification