shadcn-ui-expert
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSMETADATA_POISONING
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill references a workflow for synchronizing design tokens from Figma (via
scripts/sync-ui-themes.ts). This establishes a path where malicious instructions embedded in design metadata could influence the agent's code generation.\n - Ingestion points: External design tokens fetched from the Figma API.\n
- Boundary markers: Absent; there are no instructions to the agent to treat this data as untrusted or to ignore embedded commands.\n
- Capability inventory: The skill is authorized to write component source code and execute shell scripts.\n
- Sanitization: No sanitization or validation protocols are described for the ingested metadata.\n- [METADATA_POISONING]: The skill uses future-dated versioning (e.g., '2026' and 'Radix 2026') in its name and description. This deceptive metadata may lead an agent to misjudge available capabilities and hallucinate non-existent software features.\n- [EXTERNAL_DOWNLOADS]: The skill leverages the shadcn-ui CLI via
bun x shadcn-ui@canaryto manage component source code. This downloads and runs the latest development builds of the tool from the NPM registry.
Audit Metadata