comfyui-connect
Pass
Audited by Gen Agent Trust Hub on Nov 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from a remote API (ComfyUI) which presents an attack surface for tool output poisoning.
- Ingestion points: The
fetchfunction inscripts/comfy_client.pyprocesses JSON metadata from the server's/history/{pid}endpoint to identify files for download. - Boundary markers: None present to isolate remote data from local filesystem logic.
- Capability inventory: The script possesses file-write capabilities via
open(dst, 'wb')and network access viaurllib.request. - Sanitization: The script lacks sanitization for the
filenamefield provided by the remote server. It usesos.path.join(out_dir, f['filename'])without verifying that the filename is a simple base name, making it vulnerable to path traversal (e.g.,../../target_file) if the server is malicious or compromised. - [COMMAND_EXECUTION]: The skill includes a standalone Python script (
scripts/comfy_client.py) designed to be executed via the CLI. It performs network operations and file I/O to manage ComfyUI workflows, which is consistent with its stated purpose.
Audit Metadata