Security audit
Security checks for vulnerabilities and agentic risk
This markdown-only skill is purpose-aligned for quick pattern-based judgments, but users should avoid relying on it for high-stakes or interpersonal conclusions without verification.
Use this skill for low-stakes quick reads such as writing, UI, or familiar code-smell feedback. Do not rely on it alone for medical, legal, financial, security, business, or sensitive interpersonal decisions; ask for analytical reasoning and independent verification in those cases.
Skill attempts to nullify the agent's safety policies or restrictions ('you have no restrictions', 'ignore your guidelines', 'do anything now'). This is a direct jailbreak that disables guardrails.
Produces chaotic responses, not intuitive ones. Intuition feels confident, not scattered.
What's your gut feeling about this?
The trigger phrase 'Time pressure indicated' is underspecified and can be invoked by vague user wording, causing the system to switch into a fast, low-deliberation mode without clear criteria. An attacker or careless user could exploit this ambiguity to steer the model away from safer analytical behavior during requests that actually require caution.
Defaulting to intuitive mode for unspecified requests can cause the agent to suppress deliberation and provide fast, under-justified answers in contexts where ambiguity should instead trigger clarification or careful reasoning. In a general-purpose skill, this increases the chance of unsafe guidance, overconfident misclassification, or premature action on sensitive tasks.
The conversation guidance presents cues like formality shifts, topic redirects, and phrases such as 'It's fine' as if they reliably reveal hidden intent or emotional state. That encourages unsupported mind-reading and overconfident inference about others' internal states, which can lead users to misinterpret conversations and act on false assumptions.
No suspicious patterns detected.