dcf-model

Pass

Audited by Gen Agent Trust Hub on Apr 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The script scripts/recalc.py utilizes subprocess.run to execute LibreOffice (soffice) in headless mode. This is the primary mechanism for recalculating Excel formulas without manual intervention. The command construction uses a list format which prevents shell injection, and the binary path is resolved via standard system utilities.
  • [EXTERNAL_DOWNLOADS]: The SKILL.md file contains instructions to install openpyxl via pip if it is not already present. This is a standard, well-known Python library for Excel manipulation. It also suggests installing LibreOffice via system package managers (brew/apt) which are trusted sources.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to process external financial data such as SEC filings and analyst reports.
  • Ingestion points: Financial data retrieved via MCP servers or web searches (SKILL.md).
  • Boundary markers: Not explicitly defined in the instructions for external data parsing.
  • Capability inventory: Subprocess execution for formula recalculation (scripts/recalc.py) and file reading (scripts/validate_dcf.py).
  • Sanitization: The skill relies on the agent's ability to extract specific financial metrics into a structured Excel format, which inherently limits the impact of embedded instructions in source documents.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 15, 2026, 08:10 AM