aistudio-apps-warehouse-operations-shortages
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests artifact metadata from local and remote sources through the aistudio tool, creating an attack surface for indirect prompt injection. 1. Ingestion points: Local workspace files and read-only connected environment lookups via the aistudio skill. 2. Boundary markers: The skill relies on structured interaction flows and mandatory user approval gates. 3. Capability inventory: The skill has the capability to create and modify files and execute validation commands through the aistudio tool. 4. Sanitization: Risk is mitigated by a mandatory human-in-the-loop approval step for all proposals and by restricting environment lookups to read-only mode.
- [COMMAND_EXECUTION]: The skill performs discovery, inspection, and validation using the aistudio tool. These operations are restricted to the guided build process for warehouse apps and are subject to user review and approval before any material changes are applied to the local project.
Audit Metadata