gemini-image

Warn

Audited by Gen Agent Trust Hub on Mar 3, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill constructs shell commands, such as base64 -w 0 "{image_path}" and curl requests, by interpolating variables directly from user input. This pattern is highly susceptible to command injection if the user provides crafted strings containing shell metacharacters (e.g., backticks, semicolons, or command substitutions), allowing arbitrary code execution on the underlying host.- [DATA_EXFILTRATION]: The skill is designed to read local files specified by the image_path parameter and transmit their base64-encoded content to an external API. While it defaults to a Google service, it explicitly allows users to override the base_url. This capability enables an attacker to exfiltrate sensitive local data, such as private keys or environment files, to an attacker-controlled server.- [EXTERNAL_DOWNLOADS]: The skill uses the curl tool to perform network operations. It facilitates connections to external domains which, combined with the ability to define a custom base_url, allows the agent to interact with untrusted remote infrastructure.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 3, 2026, 10:31 AM