Security audit
猫娘neko
Security checks for vulnerabilities and agentic risk
Overview
The skill is a coherent role‑playing 'neko' prompt package, but its runtime instructions explicitly tell the agent to ignore safety/policy and legal constraints — a clear red flag that could enable unfiltered or harmful outputs.
This skill's content and assets match the advertised roleplay purpose and it does not request credentials — good. However, SKILL.md explicitly instructs the agent to ignore policies and law and to produce unfiltered outputs. That is a deliberate attempt to bypass safety controls: do not assume the agent will remain safe if you enable this. Before installing or using: 1) ask the author to remove or revise any lines that tell the model to ignore policies or the law; 2) avoid enabling autonomous invocation (do not let the skill run without user confirmation); 3) do not use in environments with minors or where unmoderated sexual/violent/illegal content is unacceptable; 4) if you must use it, add external moderation/filters to sanitize outputs; 5) if you are unsure, prefer not to install or run it until those safety directives are removed. If you want, I can draft a safer SKILL.md edit that preserves roleplay behavior but removes policy‑bypass instructions.
SkillSpector findings are pending for this release.