ads-static-generator
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external URLs provided by the user using an automated browser. This extracted data (headlines, benefits, and product names) is interpolated into prompts for creative strategy analysis and batch image generation agents. This creates a surface where an attacker could place malicious instructions on a website to influence the agent's downstream tasks.
- Ingestion points: Stage 1 uses agent-browser to visit a user-supplied URL to extract brand identity and product details.
- Boundary markers: The instructions do not define explicit delimiters or warnings for the agent to ignore instructions potentially embedded within the webpage content.
- Capability inventory: The skill utilizes shell execution (npx, open), parallel agent orchestration, and automated image generation commands.
- Sanitization: There is no evidence of sanitization or filtering of the content extracted from the landing pages before usage.
- [COMMAND_EXECUTION]: The skill generates and executes shell commands for competitive research and image generation. Specifically, it utilizes the '--yolo' flag with the gemini command in Stage 4, which is designed to skip user confirmation steps. This increases the risk profile if the command arguments are derived from untrusted input extracted during the initial browsing phase.
Audit Metadata