creative-image-generator

Fail

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: HIGHPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [Indirect Prompt Injection] (HIGH): The skill follows a workflow that reads untrusted external data which can influence its behavior.
  • Ingestion points: The skill explicitly reads local files at .claude/ads-agent/config/briefs/{account}.md (Step 2).
  • Boundary markers: There are no delimiters or instructions provided to the agent to ignore any embedded commands or adversarial instructions within these brief files.
  • Capability inventory: The skill utilizes the generate_creative_image and expand_to_stories MCP tools, which result in file creation on the local system.
  • Sanitization: No validation or sanitization of the content from the brief files is performed before it is used to populate tool parameters.
  • [Unverifiable Dependencies] (LOW): The skill requires the installation of the google-generativeai Python package. While this is a trusted package from a verified organization (Google), the request to install external code at runtime is noted. Per [TRUST-SCOPE-RULE], this is downgraded to LOW/INFO.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 16, 2026, 07:17 AM