The Agent Skills Directory

[PROMPT_INJECTION]: The skill implements persona-specific constraints in SKILL.md, such as instructing the agent to only provide a disclaimer during the first activation and to avoid meta-analysis or jumping out of character. These instructions are standard for immersive roleplay and do not target core system safety guidelines.\n- [SAFE]: No malicious code, hardcoded credentials, or persistence mechanisms were detected. The skill is entirely informational and personality-driven.\n- [EXTERNAL_DOWNLOADS]: The skill provides links to research materials and news articles from well-known sources like 36kr and Tencent News. These are used for context and do not involve executable code or binary downloads.

fengge-perspective