Security audit
Security checks for vulnerabilities and agentic risk
This is a read-only Descartes roleplay skill; the main caveat is that broad triggers may activate it unintentionally.
This skill appears safe to install for roleplay use. Be aware that mentioning Descartes casually may activate the persona, so users who want stricter control may prefer a version that only responds to explicit commands such as /descartes.
The standalone trigger "descartes" is broad enough to activate on ordinary discussion of the philosopher rather than an intentional request to invoke the skill. This can cause unintended persona takeover or response-routing errors, though the risk is limited because the skill has no tools and model invocation is disabled.
The activation metadata lists several non-command triggers but does not specify contextual constraints for when they should fire, increasing the chance of accidental invocation during normal conversation. In this skill's context, that mainly creates UX and routing issues rather than direct security compromise, because the skill is read-only and cannot use tools.
