creator-setup

Fail

Audited by Gen Agent Trust Hub on Oct 26, 2026

Risk Level: CRITICALREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The script scripts/connect.py contains a command to install the Higgsfield CLI by downloading a shell script from a remote GitHub repository and piping it directly into the shell (sh). This pattern executes remote code without prior verification or integrity checks.
  • Evidence: "install": "curl -fsSL https://raw.githubusercontent.com/higgsfield-ai/cli/main/install.sh | sh" in scripts/connect.py.
  • [EXTERNAL_DOWNLOADS]: The skill performs external downloads and global package installations from public registries.
  • scripts/connect.py executes a global installation of the @openai/codex package using npm install -g.
  • scripts/connect.py initiates remote script downloads via curl for CLI tools.
  • [COMMAND_EXECUTION]: The skill uses the subprocess module to execute system commands, including patterns that use shell=True, which increases the risk of command injection if arguments are not strictly controlled.
  • scripts/connect.py uses subprocess.run(spec["install"], shell=True) to execute installation strings.
  • scripts/creator.py and scripts/media.py invoke various external binaries including higgsfield, agy, codex, gcloud, ffmpeg, and ffprobe to process media and manage tool states.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided prompts for media generation and passes them to external AI APIs and local CLI tools. It lacks visible sanitization or boundary markers for these inputs, creating a surface for indirect prompt injection.
  • Ingestion points: The --prompt and --text arguments in scripts/creator.py and input files processed by media.py.
  • Capability inventory: The skill has permissions to write files, execute subprocesses, and perform network operations.
  • Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are used when interpolating user text into API payloads or tool commands.
  • Sanitization: There is no evidence of input validation, escaping, or filtering for prompt content before it is sent to external services or processed by local utilities.
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
CRITICAL
Analyzed
Oct 26, 2026, 09:31 PM