tanstack-start
Pass
Audited by Gen Agent Trust Hub on Mar 6, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [COMMAND_EXECUTION]: In
flows/setup/FLOW.md, the skill instructs the agent to execute a shell command incorporating a user-provided URL (<preset-url>) viabunx --bun shadcn@latest create. This pattern is susceptible to command injection if the input is not strictly validated or sanitized by the agent. - [EXTERNAL_DOWNLOADS]: The skill triggers numerous package installations from the NPM registry and involves downloading configuration from a dynamic user-specified URL during the shadcn/ui setup phase.
- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface (Category 8) by ingesting untrusted data (the shadcn preset URL) and using it in a command-line capability. Evidence: 1. Ingestion point:
flows/setup/FLOW.mdStep 14 (user input). 2. Boundary markers: Absent. 3. Capability inventory:bunxsubprocess call. 4. Sanitization: Absent.
Audit Metadata