find-the-original-image
Pass
Audited by Gen Agent Trust Hub on Aug 24, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: The
reference/preprocessing.mdfile provides numerous shell command templates for utilities includingmagick(ImageMagick),exiftool,ffmpeg,tesseract, andyt-dlp. These commands allow the agent to perform complex file manipulations and metadata extraction. - [PROMPT_INJECTION]: The skill facilitates the processing of untrusted data from the open web, creating a potential surface for indirect prompt injection.
- Ingestion points: media files provided for analysis, text strings extracted via OCR (Google Lens/Tesseract), and the textual content of match pages retrieved from external search engines (Yandex, Bing, etc.).
- Boundary markers: Absent. The instructions do not specify the use of delimiters or provide instructions to the agent to disregard commands potentially embedded in the processed data.
- Capability inventory: The skill provides the agent with the ability to execute subprocesses for a wide range of powerful media and download tools as described in
reference/preprocessing.md. - Sanitization: Absent. The provided command templates do not demonstrate the use of escaping or validation for variables such as filenames or extracted text strings.
- [EXTERNAL_DOWNLOADS]: The skill directs the agent to interact with various third-party search engines and provides instructions for using
yt-dlpto download video content from remote platforms for keyframe analysis.
Audit Metadata