ugc-content-factory
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input including product names, audience descriptions, and external URLs, which are interpolated into prompts for fal.ai and Nano Banana Pro models without sanitization or boundary markers. * Ingestion points: Product context and image URLs defined in SKILL.md and references/CHARACTER_LIBRARY.md. * Capability inventory: The skill makes external network requests and executes shell commands via test scripts. * Boundary markers: None present. * Sanitization: None present.
- [DATA_EXFILTRATION]: The skill includes instructions and scripts to perform network requests to fal.run (fal.ai API), which is an external domain not included in the predefined whitelist.
- [COMMAND_EXECUTION]: The skill contains a shell script (test-fal-api.sh) that executes curl and jq commands to interact with the fal.ai API.
Audit Metadata