Security audit
Security checks for vulnerabilities and agentic risk
This security-assessment skill is not malicious, but it asks for broad command execution and live-target assessment capabilities without tight authorization or data-flow limits.
Use this only for systems you own or are explicitly authorized to assess. Review each command, URL target, callback URL, and API key before use, and prefer a sandboxed environment with limited access to secrets and production systems.
The skill claims unauthorized penetration testing is out of scope, but it simultaneously advertises scanning live URLs and exposes the exec tool. That mismatch can mislead users and downstream agents into performing active security actions against external targets under the guise of benign assessment, increasing the risk of unintended offensive use.
The invocation language is broad and ambiguous, such as using the skill whenever security checks, compliance audits, vulnerability scanning, or encryption protection are needed. Such generic triggers increase the chance that an orchestrating agent invokes this exec-capable skill for routine requests where safer, narrower tools should be used, raising the risk of overreach and unintended actions.
The documented purpose is a cybersecurity assessment engine, but later sections broaden capabilities to generic file handling, API integration, and system command execution. This expands the operational scope beyond user expectations and enables the skill to act as a general-purpose execution wrapper, which is dangerous in an agent environment because it can be repurposed for unrelated or harmful actions.
Advertising file writing, external API access, and command execution without strong user-facing warnings hides the real-world impact of using the skill. In an agent context, these capabilities can modify local state, exfiltrate data to remote services, or run destructive commands if triggered by ambiguous prompts or hostile input.
