deepfake-detection

Pass

Audited by Gen Agent Trust Hub on Apr 15, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill provides instructions for downloading the c2patool utility from the Content Authenticity Initiative's official GitHub repository and uses system package managers like Homebrew, APT, and WinGet for installing well-known forensic tools including ffmpeg, exiftool, and imagemagick.
  • [COMMAND_EXECUTION]: The agent is instructed to use standard command-line interfaces for ffmpeg, exiftool, and magick to extract signal data and metadata from media files, which are routine operations for forensic verification.
  • [PROMPT_INJECTION]: This skill possesses an indirect prompt injection surface due to its primary function of processing untrusted external media files. Ingestion points: Media files are ingested via tools like ffmpeg and exiftool in SKILL.md. Boundary markers: No specific delimiters or instructions to ignore embedded commands are provided for tool outputs. Capability inventory: The agent can execute shell commands (ffmpeg, exiftool, magick) and perform file system reads. Sanitization: While structured output parsing is mentioned, there is no explicit sanitization of malicious instructions that could be hidden in media metadata or visual content.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 15, 2026, 05:26 PM