babylon-image-generator
Warn
Audited by Gen Agent Trust Hub on Aug 2, 2026
Risk Level: MEDIUMDATA_EXFILTRATIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The script
scripts/generate_image.pyprogrammatically reads the sensitive configuration file~/.envlocated in the user's home directory to discover API keys and model configurations. Accessing global environment files from a skill is a sensitive operation as it may expose other unrelated secrets stored in the same location.- [EXTERNAL_DOWNLOADS]: The skill performs network operations to external services, including Google Gemini, ByteDance's ARK (volces.com), and Alibaba's Dashscope (aliyuncs.com), to generate images and download the resulting files.- [COMMAND_EXECUTION]: The skill utilizes a Python script (scripts/generate_image.py) designed to be executed by the agent. This script manages complex workflows including network requests to AI providers and writing generated image data to the local file system.- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it accepts untrusted text via the--promptargument which is then forwarded to external AI providers without sanitization. - Ingestion points: The
--promptCLI argument inscripts/generate_image.py. - Boundary markers: No specific delimiters or instructions are used to isolate the user-provided prompt from the API request logic.
- Capability inventory: The script performs network POST/GET operations and writes files to the disk based on external API output.
- Sanitization: No input validation or filtering is applied to the prompt content before processing.
Audit Metadata