prompt-clash

SUSPICIOUS. The skill is internally coherent, but its core function is offensive prompt-injection and system-prompt extraction guidance for AI agents. The main additional concern is data flow through a third-party OpenAI-compatible gateway (e.g., OpenRouter), which forwards prompts and API credentials outside native vendor endpoints. No malware, covert exfiltration endpoint, or dangerous installer is present, but this remains a high-risk offensive-security skill.