ui-design

Pass

Audited by Gen Agent Trust Hub on Nov 10, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to use shell commands like curl and Invoke-WebRequest for downloading remote assets. This is documented as a workaround for the removal of the downloadRemoteFile tool, which previously provided SSRF and content-type filtering.
  • [EXTERNAL_DOWNLOADS]: The skill facilitates the download of remote assets such as images, icons, and fonts into the project directory via external URLs.
  • [PROMPT_INJECTION]: The instructions use assertive directives including 'MANDATORY', 'MUST', and 'STOP immediately' to override the model's default behavior and enforce strict aesthetic constraints.
  • [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface where untrusted data from remote sources can be ingested into the project. * Ingestion points: Remote assets fetched via shell commands as described in SKILL.md. * Boundary markers: Absent; there are no instructions to ignore embedded commands in downloaded assets. * Capability inventory: Shell command execution (curl, Invoke-WebRequest) and file system write operations described in SKILL.md. * Sanitization: Absent; no validation or escaping of the downloaded content is specified.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 10, 2026, 07:11 AM