21st-ui-build
Pass
Audited by Gen Agent Trust Hub on Sep 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute multiple shell commands using the
21stCLI to manage the UI development lifecycle. - Commands include:
21st init --design-context,21st search,21st usage,21st generate, and21st review. - [EXTERNAL_DOWNLOADS]: The workflow involves fetching external UI components and inspiration via
21st searchand21st getcommands. - These tools interact with the 21st-dev ecosystem to retrieve component data and implementation patterns.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from external search results and local project configuration files which may influence the agent's implementation decisions.
- Ingestion points: Reads project configuration in
.21st/design.jsonand.21st/DESIGN.md, and parses results from the21st searchcommand. - Boundary markers: None identified in the instructions to separate untrusted data from system instructions.
- Capability inventory: Includes shell command execution (21st CLI) and file system writes (storing design decisions).
- Sanitization: None identified; the agent is instructed to trust explicit project facts and search results.
Audit Metadata