Security audit
Security checks for vulnerabilities and agentic risk
This is a prompt-rewriting skill with no code execution, persistence, credential handling, or external access in the inspected artifact.
Before installing, understand that the skill may rewrite vague prompts based on its own interpretation. Review the optimized prompt before using it, especially for legal, medical, financial, security, or other sensitive tasks.
The instruction to automatically optimize vague input according to the 'most reasonable intent' permits the skill to infer goals that the user did not clearly specify. This can lead to unauthorized semantic changes, accidental escalation of risky requests, or optimization of harmful content in a more effective form.
The skill claims applicability across many domains and lacks explicit trigger boundaries or eligibility checks, which can cause it to activate on inputs outside its intended scope. In prompt-handling systems, overly broad scope increases the chance of unintended transformation of sensitive, safety-critical, or high-stakes user prompts without user confirmation.
The instruction says the skill will automatically detect Chinese or English and keep the language consistent, but it does not indicate that users may override or choose output language. This can be a natural-language policy concern if organizational policy requires language or locale choice rather than implicit enforcement.
No suspicious patterns detected.
