grok-imagine-image
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEDATA_EXFILTRATIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill performs network requests to an external IP address (43.163.230.83:8000) for API communication and image retrieval. This endpoint is not a well-known service or a whitelisted domain.
- [EXTERNAL_DOWNLOADS]: The included Python script downloads generated images from the remote API using the urllib.request module.
- [INDIRECT_PROMPT_INJECTION]: The skill acts as a surface for indirect prompt injection due to its handling of untrusted data.
- Ingestion points: User prompts are accepted via CLI arguments (--prompt) or text files (--prompt-file) in SKILL.md, and structured data is ingested from the external API's JSON response in scripts/generate.py.
- Boundary markers: No specific delimiters or instructions are used to prevent the agent from following potential commands embedded within the prompts or API responses.
- Capability inventory: The skill utilizes network capabilities (urllib.request) and file system access (Path.write_bytes).
- Sanitization: While the script uses basic slugification for filenames and standard JSON encoding for payloads, it does not sanitize the content of inputs or API responses for injection attacks.
Audit Metadata