design-lab
Warn
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the agent to fetch and install external code from the
0xdesign/design-pluginrepository on GitHub. This is performed vianpx skills addand specialized plugin marketplace commands. - [COMMAND_EXECUTION]: The skill provides shell commands for the agent to verify installation status using
grepand file system checks, and to perform installation usingnpx. - [INDIRECT_PROMPT_INJECTION]: The skill has a vulnerability surface for indirect prompt injection as it processes external feedback to refine code.
- Ingestion points: Element-level feedback collected through a browser overlay (
SKILL.md). - Boundary markers: None identified in the provided instructions to separate external feedback from agent instructions.
- Capability inventory: The skill generates and writes new UI components (code) to the user's filesystem and removes temporary files.
- Sanitization: No sanitization or validation of the feedback content is mentioned before it is synthesized into refined code.
- [DYNAMIC_EXECUTION]: The skill generates "real working components" (executable UI code) at runtime based on its design exploration and feedback synthesis loop.
Audit Metadata