design-review-package
Pass
Audited by Gen Agent Trust Hub on Nov 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill operates entirely on local files provided by the user or bundled with the skill. The logic for processing FMEA and BOM data is contained within a transparent Python script using standard libraries without any external network dependencies or suspicious execution patterns.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes external data (user-provided CSV files) for reporting purposes, which presents a typical ingestion surface. \n
- Ingestion points: The
scripts/review_tables.pyscript readsfmea.csvandbom.csvprovided by the user.\n - Boundary markers: No explicit instructions to ignore embedded instructions are present in the script or prompt.\n
- Capability inventory: The script has file-write capabilities (generating
.csvand.mdreports) and is executed via the agent's shell command capability.\n - Sanitization: Numeric fields are validated via type casting, but string fields (e.g., descriptions, effects) are not escaped before being written into markdown tables in the output reports.
Audit Metadata