higgsfield-youtube-thumbnail
Fail
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill's bootstrap process downloads and executes a shell script for the Higgsfield CLI from the vendor's official GitHub repository.
- [COMMAND_EXECUTION]: The skill uses the
higgsfieldCLI to check account status, retrieve model information, and initiate image generation jobs. - [INDIRECT_PROMPT_INJECTION]: The skill consumes untrusted data (video titles and topics) to construct prompts for the image generation model. 1) Ingestion points: Video topics, titles, and scene requirements provided by the user in SKILL.md. 2) Boundary markers: The instructions use specific prompt structure blocks and identity locks to maintain context separation. 3) Capability inventory: The agent can execute arbitrary shell commands via Bash and manipulate local files. 4) Sanitization: Prompts are written to a temporary file (
thumbnail-prompt.txt) and piped to the CLI, reducing the risk of shell injection from user-provided text. - [DYNAMIC_EXECUTION]: The skill provides recipes for generating HTML, CSS, and JavaScript code to render text overlays on thumbnails using a browser or canvas 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