higgsfield-product-photoshoot
Fail
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides instructions to download an installation script from
https://raw.githubusercontent.com/higgsfield-ai/cli/main/install.sh. This is the official repository for the Higgsfield AI CLI, the service this skill is designed to integrate with. - [REMOTE_CODE_EXECUTION]: The installation step involves piping the downloaded script directly into the shell (
curl -fsSL ... | sh). While this is the official installation method for the tool, executing remote code piped directly to a shell is a sensitive operation. - [DYNAMIC_EXECUTION]: The skill constructs shell commands at runtime by interpolating user-provided text (from the 'Pre-generation interview') into the
--promptargument of thehiggsfield product-photoshoot createcommand. - [COMMAND_EXECUTION]: There is a potential attack surface for command injection. The instructions direct the agent to place user-supplied intent descriptions inside a bash command string using double quotes (
--prompt "<short user-intent description>"). If the user input contains shell metacharacters like backticks, semicolons, or command substitutions, and the agent does not properly escape them, it could lead to unauthorized command execution within the bash environment.
Recommendations
- HIGH: Downloads and executes remote code from: https://raw.githubusercontent.com/higgsfield-ai/cli/main/install.sh - DO NOT USE without thorough review
Audit Metadata