The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes shell and Python scripts for content analysis.
fetch_youtube_subtitles.sh: Downloads YouTube subtitles using yt-dlp. It installs yt-dlp via pip if not found. While it executes shell commands, it is a legitimate tool within the stated purpose of analyzing video content.
analyze_titles.py, retention_curve_checker.py, thumbnail_audit.py: These are pure text-processing scripts for analyzing video metadata and scripts. They perform no network operations or sensitive file access.
[REMOTE_CODE_EXECUTION]: The README mentions npx skills add alchaincyf/mrbeast-skill. This uses the official vercel-labs/skills installer, which is a recognized service for managing agent skills. The installation command is standard for this ecosystem.
[DATA_EXPOSURE]: No hardcoded credentials, sensitive file paths, or unauthorized network exfiltration patterns were detected in the instructions or scripts.
[PROMPT_INJECTION]: The skill uses a role-play prompt to adopt the 'MrBeast' persona. It includes clear 'EXIT TRIGGER' phrases ('退出', '切回正常') to allow the user to escape the persona, which is a safety best practice. It does not attempt to bypass underlying AI safety guardrails.

mrbeast-perspective