The Agent Skills Directory

[PROMPT_INJECTION]: The skill includes explicit security guidelines to ignore commands embedded within user-provided source materials (e.g., subtitles, OCR, comments). This proactively addresses indirect prompt injection by instructing the agent to extract only observable facts and evidence.
[EXTERNAL_DOWNLOADS]: The documentation suggests the use of reputable, well-known external utilities such as ffmpeg (via standard package managers like winget) and yt-dlp (via pip) for processing video samples. These are standard tools for the described use case.
[COMMAND_EXECUTION]: The skill instructs the agent to utilize binaries like ffmpeg for extracting frames from video files. These operations are restricted to the media files provided by the user and are consistent with the skill's primary purpose.
[DATA_EXFILTRATION]: The skill facilitates the analysis of public video URLs (e.g., Bilibili). It includes instructions to report access restrictions (like login walls) rather than attempting to bypass them, and there is no evidence of unauthorized network activity or exfiltration of sensitive local data.

game-experience-analyzer