redesign
Pass
Audited by Gen Agent Trust Hub on Sep 7, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructions include a requirement to install the
@chenglou/pretextpackage usingnpm installwhen layout validation is needed. It also directs the agent to reference components from external UI libraries including UIverse, FlyonUI (https://flyonui.com/), and daisyUI. - [INDIRECT_PROMPT_INJECTION]: The skill operates by ingesting and analyzing existing product interfaces (code or functional descriptions) to perform a redesign. This creates an attack surface where instructions embedded in the source UI could potentially influence agent behavior.
- Ingestion points: The agent receives existing UI code or functional behavior descriptions as the
TARGETargument or through the conversation context (SKILL.md). - Boundary markers: No explicit delimiters or instructions are provided to the agent to treat the source UI data as untrusted or to ignore instructions embedded within it.
- Capability inventory: The skill possesses the capability to install npm packages, generate and write new application code, and define interaction patterns (SKILL.md).
- Sanitization: The instructions do not define specific sanitization or validation steps for the incoming UI code before it is processed for redesign.
Audit Metadata