media-gpt-image2-builder
Pass
Audited by Gen Agent Trust Hub on May 14, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by interpolating user-provided natural language into an internal 8-step prompt template.
- Ingestion points: Untrusted data enters via the
user_promptfield inSKILL.mdand is processed by the logic inreferences/8-step-prompt.md. - Boundary markers: The skill lacks explicit delimiters (such as XML tags or triple quotes) to isolate the user input from the internal prompt instructions.
- Capability inventory: The skill has the capability to call the OpenAI image generation API (
v1/images/generations) and chain with themedia-ai-disclosureskill. - Sanitization: There is no evidence of input validation or sanitization for the
user_promptbefore it is used in the image generation process. - Remediation: To mitigate risks, wrap user input in distinct delimiters and include explicit instructions for the agent to treat the content within those delimiters as data only, ignoring any embedded commands.
Audit Metadata