The Agent Skills Directory

[SAFE]: The skill operates as a content production coordinator with a focus on non-fiction media. It enforces strict 'no-fabrication' rules and requires human-in-the-loop verification for platform data (e.g., manually searching for trending titles rather than automated scraping).
[INDIRECT_PROMPT_INJECTION]: The skill processes external materials (user files, pasted text, screenshots). However, it includes a dedicated safety protocol (external-material-safety) to mitigate risks associated with processing third-party content. It also specifically warns against deceptive practices like plagiarism ('洗稿') or falsifying personal experiences.
[COMMAND_EXECUTION]: All referenced tools and files are accessed through local relative paths within the skill's defined environment (e.g., ../eva-shared/). There are no attempts to execute unauthorized system commands or external scripts.
[DATA_EXFILTRATION]: No network operations (curl, wget, etc.) or hardcoded credentials were found. Data flow is restricted to internal asset management and user interaction within the agent's context.

eva-create