image-to-3d

Warn

Audited by Gen Agent Trust Hub on Nov 4, 2026

Risk Level: MEDIUMDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [DYNAMIC_EXECUTION]: The fetch function in scripts/comfy_client.py performs unsafe file path construction by joining a local output directory with a filename provided directly by the remote ComfyUI server without any validation. This creates a path traversal vulnerability where a malicious server can specify a filename containing traversal sequences (e.g., ../../) to overwrite sensitive files outside of the intended directory.
  • [INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection via tool output poisoning because it ingests structured data from an external API and uses it to perform file system operations. Ingestion points: API responses from the ComfyUI server in scripts/comfy_client.py. Boundary markers: None identified in the script logic to delimit or warn about untrusted content. Capability inventory: The skill possesses file writing capabilities and network access to the ComfyUI API. Sanitization: The script fails to sanitize or validate filenames received from the remote server before using them in file path construction.
  • [DATA_EXFILTRATION]: The client script handles a sensitive COMFY_API_KEY environment variable and transmits it via an Authorization header to the address specified in the COMFY_URL environment variable. If an attacker can influence the value of COMFY_URL to point to a malicious server, the API key will be exfiltrated to that external domain.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Nov 4, 2026, 08:11 PM