gpt-image-2-prompt-engine
Pass
Audited by Gen Agent Trust Hub on Sep 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user data via parameters like subject and text_content which are interpolated into image generation prompts. The ingestion points are located in the main workflow of SKILL.md. While the skill lacks explicit boundary markers or sanitization for this data, its capabilities are limited to the image_generation tool and local template searching, presenting a low-risk surface inherent to the skill's purpose as a prompt engine.
- [EXTERNAL_DOWNLOADS]: The skill references a third-party GitHub repository (freestylefly/awesome-gpt-image-2) and associated raw content links for documentation and template retrieval. These references are used legitimately to support the prompt engineering methodology and do not involve executing untrusted code.
- [COMMAND_EXECUTION]: The workflow involves running a local Python utility script (scripts/query_templates.py) to query design templates. Technical review of the script confirms it only performs local file reads and prints informational URLs without performing network operations or accessing sensitive credentials.
Audit Metadata