gemini-image-generator
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSDATA_EXFILTRATION
Full Analysis
- [EXTERNAL_DOWNLOADS] (LOW): The script
scripts/recraft_process.pydownloads content from URLs provided by the Recraft API response and writes them to local files. - [DATA_EXFILTRATION] (LOW): The skill transmits local image files to the Recraft API (
external.api.recraft.ai) for processing (background removal and vectorization). While this is the intended functionality, it involves sending data to a non-whitelisted third-party domain. - [REMOTE_CODE_EXECUTION] (SAFE): No dynamic code execution (eval/exec) or piped remote script execution was detected. The skill uses standard API clients.
- [CREDENTIALS_UNSAFE] (SAFE): API keys are correctly managed via environment variables (
GEMINI_API_KEY,RECRAFT_API_KEY). No hardcoded secrets were found. - [INDIRECT_PROMPT_INJECTION] (LOW): The skill ingests untrusted user input via the
--promptargument and interpolates it directly into the Gemini API call. - Ingestion points:
args.promptinscripts/generate.py. - Boundary markers: None; input is passed as a raw string to the model.
- Capability inventory: File writing via
Image.saveandos.makedirs. - Sanitization: None detected.
Audit Metadata