fal-ai-media
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the user to configure an MCP server by running
npx -y fal-ai-mcp-server. This fetches and executes the official MCP server package from the npm registry. - [COMMAND_EXECUTION]: The setup instructions involve executing a shell command via
npxto initialize the fal.ai MCP server. This is a standard procedure for integrating MCP tools. - [DYNAMIC_EXECUTION]: The skill provides Python code snippets that use the
requestslibrary to interact with external APIs (ElevenLabs). These snippets demonstrate safe practices by retrieving sensitive credentials from environment variables rather than hardcoding them. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input to generate AI media.
- Ingestion points: User-provided text strings are interpolated into the
promptandtextparameters for image, video, and audio generation tools (SKILL.md). - Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat the user prompts as data rather than instructions.
- Capability inventory: The skill allows for generating media files and uploading data to the fal.ai service, but it does not have the capability to write to the local file system (beyond a specific Python example) or execute arbitrary system commands based on tool output.
- Sanitization: No content filtering or validation is described for the user-supplied prompts before they are sent to the generation models.
Audit Metadata