Skills
▥
skills/inference-sh-9/skills/nano-banana-2/Gen Agent Trust Hub

nano-banana-2

Fail

Audited by Gen Agent Trust Hub on Feb 26, 2026

Risk Level: HIGHEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTIONREMOTE_CODE_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches the infsh CLI installation script from the vendor's official domain (cli.inference.sh).
  • [COMMAND_EXECUTION]: The skill uses the Bash tool to execute infsh commands. This is restricted to the vendor's specific tool via the infsh * pattern.
  • [REMOTE_CODE_EXECUTION]: The installation instructions include a curl | sh pattern directed at the vendor's own infrastructure (https://cli.inference.sh). This is the standard deployment method for this vendor's tools.
  • [PROMPT_INJECTION]: The skill exposes an indirect prompt injection surface by allowing the ingestion of untrusted data from external sources.
  • Ingestion points: The images array in the input JSON allows users to provide external URLs that are processed by the Gemini model via the infsh tool.
  • Boundary markers: No explicit boundary markers or instructions to ignore embedded content are present in the command templates.
  • Capability inventory: The skill has the capability to execute the infsh CLI tool via Bash subprocesses.
  • Sanitization: There is no evidence of sanitization or validation of the remote image content before it is passed to the model.
Recommendations
  • HIGH: Downloads and executes remote code from: https://cli.inference.sh - DO NOT USE without thorough review
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 26, 2026, 03:13 PM