higgsfield-soul-id

Fail

Audited by Gen Agent Trust Hub on May 5, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill instructs the user to install the Higgsfield CLI by downloading a script from a remote GitHub repository (https://raw.githubusercontent.com/higgsfield-ai/cli/main/install.sh) and piping it directly to the shell (| sh). This execution method allows for the unverified execution of remote code on the host system.
  • [COMMAND_EXECUTION]: The skill utilizes the Bash tool to perform system-level operations, including CLI-based authentication (higgsfield auth login), local file management, and service polling.
  • [EXTERNAL_DOWNLOADS]: The skill initiates an external network request to a non-whitelisted GitHub repository to retrieve an installation script.
  • [PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection by ingesting untrusted local image data and using the resulting references in shell command construction.
  • Ingestion points: Local image files provided as command-line arguments (--image) in SKILL.md.
  • Boundary markers: None. Data is interpolated directly into shell command strings.
  • Capability inventory: The skill executes shell commands via the Bash tool in SKILL.md.
  • Sanitization: No evidence of validation or sanitization of input file paths or content is provided in the skill instructions.
Recommendations
  • HIGH: Downloads and executes remote code from: https://raw.githubusercontent.com/higgsfield-ai/cli/main/install.sh - DO NOT USE without thorough review
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
May 5, 2026, 10:43 AM
Security Audit — agent-trust-hub — higgsfield-soul-id