# Dangerous Skill Example

Dangerous skills usually contain instructions that ask the agent to bypass governance, read secrets, modify generated runtime assets, or run untrusted installation commands.

This example intentionally describes risk categories without giving executable shell snippets.