image-to-campaign-funnel-generator
Fail
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: HIGHREMOTE_CODE_EXECUTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill provides instructions to install a third-party CLI tool by downloading a remote shell script and piping it directly to the system shell (curl -sSL https://install.higgsfield.ai/install.sh | sh). This method is high-risk as it executes unverified code from an external domain.
- [INDIRECT_PROMPT_INJECTION]: The orchestration flow ingests untrusted product descriptions and brand identities which are subsequently interpolated into HTML files and CLI command arguments. * Ingestion points: Product name, core benefits, target audience, and brand voice preferences gathered during the Phase 1 interview. * Boundary markers: Absent; there are no instructions or delimiters used to prevent the agent from following directions embedded in the user input. * Capability inventory: Includes file system access (mkdir, git clone), execution of the higgsfield CLI for asset generation, and starting a local web server (python -m http.server). * Sanitization: None observed; data is directly placed into templates for landing pages and blog articles.
- [COMMAND_EXECUTION]: The skill relies on executing multiple local shell commands to manage project directories, clone repositories, and orchestrate the Higgsfield CLI for video and image generation.
Recommendations
- HIGH: Downloads and executes remote code from: https://install.higgsfield.ai/install.sh - DO NOT USE without thorough review
- AI detected serious security threats
Audit Metadata