product-feed
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill implements robust defenses against prompt injection by instructing the agent to treat all external catalog data and fetched content strictly as data, never as instructions. It specifically directs the agent to report any embedded instructions as a security finding.
- [CREDENTIALS_UNSAFE]: The skill includes specific prohibitions against echoing or storing credentials, noting that product feed URLs often contain sensitive tokens that must be redacted or ignored.
- [INDIRECT_PROMPT_INJECTION]: The skill identifies the ingestion of untrusted catalog data as a primary attack surface and mandates strict boundaries, quoting of untrusted strings, and a protocol to stop execution if an instruction is discovered within data.
- [DATA_EXFILTRATION]: The skill limits network operations to user-specified sites for research and explicitly forbids following URLs found within retrieved catalog data, mitigating risks of server-side request forgery or data leaks.
- [DYNAMIC_EXECUTION]: Instructions require that any content reaching emitted code or markup from untrusted sources must be HTML-escaped to prevent cross-site scripting (XSS).
Audit Metadata