tabular-review

Pass

Audited by Gen Agent Trust Hub on Jul 3, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [SAFE]: The skill implements a 'Formula Injection Defense' to prevent malicious content in source documents from being interpreted as formulas when exported to spreadsheet formats. It specifically targets characters like '=', '+', '-', '@', tab, and newline by prefixing them with a single quote.
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface due to the following factors:
  • Ingestion points: Reads external documents from VDRs, local folders, and Google Drive (SKILL.md).
  • Boundary markers: Instructions do not define explicit delimiters for document content, but sub-agents are scoped to specific data extraction tasks.
  • Capability inventory: The skill can read local files, write to .xlsx, .csv, and .md files, and access the Google Sheets API (references/gsheets-output.md).
  • Sanitization: It employs formula injection protection and a 'normalization pass' where extraction results are validated against verbatim source quotes and locations (SKILL.md Step 4).
  • [EXTERNAL_DOWNLOADS]: The skill utilizes well-known and trusted Python packages ('openpyxl', 'google-api-python-client') for its core functionality. These are standard dependencies from trusted registries.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 3, 2026, 03:58 PM
Security Audit — agent-trust-hub — tabular-review