openai-image-gen

Pass

Audited by Gen Agent Trust Hub on Jul 9, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The script communicates with the official OpenAI API at api.openai.com to generate images and subsequently downloads the generated media from URLs provided by the service.
  • [PROMPT_INJECTION]: The skill contains an indirect injection surface in the scripts/gen.py file, where untrusted user input is interpolated into an HTML context.
  • Ingestion points: Untrusted data enters the skill via the --prompt command-line argument.
  • Boundary markers: There are no boundary markers or instructions to treat the prompt content as literal text rather than HTML.
  • Capability inventory: The script possesses the capability to write files to the local file system (Path.write_text, write_bytes) and perform network requests (urllib.request.urlopen, urlretrieve).
  • Sanitization: While the script sanitizes filenames using a regex-based slugify function, it fails to perform any HTML escaping or sanitization on the prompt string before inserting it into the index.html gallery file.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 9, 2026, 10:45 PM
Security Audit — agent-trust-hub — openai-image-gen