web-coder
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external web content, codebases, and user requirements, which act as ingestion points for potentially untrusted data. The instructions lack explicit boundary markers or sanitization guidelines to prevent the agent from executing instructions that might be embedded within processed HTML comments or script metadata. Because the persona is granted high autonomy for development tasks (such as build execution and server configuration), the potential impact of a successful injection is elevated.
- [EXTERNAL_DOWNLOADS]: The reference documentation includes examples of fetching resources from well-known services and trusted CDNs (e.g., JSDelivr) and links to official specifications from trusted organizations like W3C and MDN. These are documented for the developer's reference and do not involve the skill itself downloading or executing remote code in an unsafe manner.
Audit Metadata