Security audit
Security checks for vulnerabilities and agentic risk
This markdown-only skill changes the agent's tone and self-check process, but it does not add code execution, data access, persistence, or external tools.
Install this only if you want the agent to add a reflective, emotionally aware tone to ordinary tasks. Avoid or disable it when you want strictly literal, task-focused responses without inferred mood or personalization.
The usage conditions are ambiguous and subjective, especially triggers like noticing 'auto-pilot' or feeling the user is 'off.' In an agent system, these soft conditions can cause unpredictable activation and encourage the model to make unsupported judgments about the user's emotional state, which may affect outputs in ways the user did not request.
The invocation guidance is broad enough that the skill could activate for almost any user request involving execution or writing, which increases the chance it will influence unrelated tasks. Because the skill explicitly encourages the agent to infer what it 'senses' about the user and let feelings inform execution, overbroad activation can introduce unnecessary personalization, speculation about user state, and inconsistent behavior across normal tasks.
No suspicious patterns detected.
